docs: link website first-agent examples map

Co-authored-by: Codex <codex@openai.com>

.github/loop/PRIORITIES.md

@@ -21,8 +21,8 @@ changes, architectural rewrites. Those go to the human.
 
 ## Work queue (ranked)
 
-1. **Add CLI examples wayfinding for first-agent paths** ([#4115](https://github.com/micro/go-micro/issues/4115)) — keep developer adoption first while the on-ramp is the explicit gap: make the maintained no-secret first-agent, debugging, and 0→hero examples discoverable from the CLI itself. Align CLI output, README, and website guide references so the scaffold → run → chat → inspect path stays copy/pasteable after install.
-2. **Stabilize plan-delegate harness against duplicate delegated notifications** ([#4118](https://github.com/micro/go-micro/issues/4118)) — the scheduled provider-conformance matrix is now in place, and its first live signal exposed an atlascloud/minimax duplicate-notify regression. Preserve the semantic “exactly one launch-readiness notification” contract with focused regression coverage so the live matrix remains a useful evaluator rather than a noisy gate.
+1. **Link examples wayfinding from website getting-started path** ([#4241](https://github.com/micro/go-micro/issues/4241)) — top adoption gap after the examples index shipped and AtlasCloud notify follow-ups were fixed: the README and CLI now point at the first-agent/0→hero map, but go-micro.dev getting-started and quickstart still need the examples index/support reference links that make the no-secret on-ramp discoverable.
+2. **Make AtlasCloud guarded delegation pass reliably** ([#4244](https://github.com/micro/go-micro/issues/4244)) — Now-phase cross-provider conformance remains important after the Minimax request-shape fallback, duplicate delegated-notification replay fixes, OpenAI-compatible text tool-call parsing, and AtlasCloud multi-step follow-up fixes shipped; keep it in queue until the live agent harness consistently observes the guarded delegate within the retry budget.
 
 _Seeded by Claude Code from the roadmap + open issues; thereafter maintained by the
 architecture-review pass._

CHANGELOG.md

@@ -16,6 +16,58 @@ next version when it ships.
 
 ## [Unreleased]
 
+### Added
+- **StreamAsk close cancellation** — agent streaming calls now cancel promptly when their runner closes, avoiding orphaned stream work. (`agent/`)
+- **Agent resume pending helper** — agent durability now has a focused helper for resuming pending checkpointed runs. (`agent/`)
+
+### Fixed
+- **Plan/delegate retry idempotency** — agent retries now preserve side-effect and notification dedupe across conformance retry paths, including completion and owner-notification edge cases. (`agent/`, `internal/harness/`)
+
+---
+
+## [6.3.17] - July 2026
+
+### Added
+- **First-agent examples CLI wayfinding** — `micro examples` now prints the maintained provider-free first-agent examples in copy/paste order. (`cmd/micro/`)
+- **0→hero CLI entrypoint** — `micro zero-to-hero` now points developers at the maintained no-secret services → agents → workflows harness and runnable examples. (`cmd/micro/`)
+- **First-agent tutorial smoke harness** — the first-agent tutorial path now has smoke coverage to keep the no-secret on-ramp runnable. (`internal/harness/`)
+- **No-secret agent debugging smoke** — the no-secret agent debugging path now has smoke coverage for the first-agent troubleshooting flow. (`internal/harness/`)
+- **Durable checkpoint resume smoke coverage** — durable agent resume after checkpointing now has focused smoke coverage. (`agent/`, `internal/harness/`)
+
+### Fixed
+- **Plan/delegate notify replays** — duplicate and replayed plan-delegate notifications are now idempotent, so resumed runs do not duplicate completed notifications. (`agent/`, `internal/harness/`)
+- **Provider conformance scheduling** — provider conformance workflow dispatches now guard their scheduling path more reliably. (`.github/workflows/`)
+- **Plan/delegate notification completion** — delegated notifications now preserve plan completion state more reliably, including duplicate, paraphrased, and delegated-owner notification paths. (`agent/`, `internal/harness/`)
+- **AtlasCloud tool fallback** — AtlasCloud built-in tool schemas and follow-up tool fallback handling now recover conformance delegate retries more reliably. (`ai/atlascloud/`, `agent/`)
+- **Agent conformance retry completion** — conformance retry prompts and completion handling are more deterministic for delegated agent runs. (`agent/`, `internal/harness/`)
+
+### Documentation
+- **First-agent quickstart numbering** — the first-agent on-ramp numbering is consistent across the README and website docs. (`README.md`, `internal/website/docs/`)
+- **First-agent inspect command** — docs now use the maintained `micro inspect agent <name>` form. (`README.md`, `internal/website/docs/`)
+- **`micro loop` quickstart wayfinding** — docs now surface the loop quickstart from the public docs index and README wayfinding. (`README.md`, `internal/website/docs/`)
+
+---
+
+## [6.3.16] - July 2026
+
+### Added
+- **No-secret agent demo CLI** — the CLI now surfaces `micro agent demo`, making the provider-free first-agent path discoverable from the installed binary. (`cmd/micro/`)
+- **First-agent recovery doctor** — first-agent recovery checks now help diagnose install, scaffold, and provider setup issues before the live agent run. (`cmd/micro/`, `internal/website/docs/guides/`)
+
+### Changed
+- **Architecture lifecycle docs** — the architecture guide now leads with the services → agents → workflows lifecycle and the first-agent on-ramp. (`internal/website/docs/architecture.md`)
+- **First-agent on-ramp** — README and website docs now lead new users through install troubleshooting, no-secret demos, the smallest first-agent example, debugging, and the 0→hero reference path in the same order. (`README.md`, `internal/website/docs/`)
+
+### Fixed
+- **Config close idempotency** — config close paths now tolerate repeated closes safely. (`config/`)
+- **OpenTelemetry child span events** — agent traces now preserve child span events more reliably. (`agent/`)
+
+### Documentation
+- **Security reporting** — security docs now route vulnerability reports through GitHub Security Advisories. (`SECURITY.md`, `internal/website/docs/`)
+- **Install troubleshooting** — the first-agent on-ramp now includes clearer install and PATH recovery guidance. (`internal/website/docs/guides/install-troubleshooting.md`)
+
+---
+
 ## [6.3.15] - July 2026
 
 ### Added

README.md

@@ -31,6 +31,7 @@ Running Go Micro in production, or building on it and want help? Paid **support,
 - [Building Agents](#building-agents) — [Plan & Delegate](#plan--delegate), [Pluggable](#batteries-included-pluggable), [Paid tools (x402)](#paid-tools-x402), [A2A](#reachable-by-other-agents-a2a)
 - [Features](#features)
 - [CLI](#cli)
+- [Autonomous improvement loop](#autonomous-improvement-loop)
 - [Multi-Service Projects](#multi-service-projects)
 - [Data Model](#data-model)
 - [AI Providers](#ai-providers)
@@ -91,19 +92,40 @@ walkable agent path in this order:
 
 1. [Install troubleshooting](internal/website/docs/guides/install-troubleshooting.md) — verify the binary installer or `go install`, `PATH`, `micro --version`, and the no-secret smoke path before agent work.
 2. `micro agent demo` — print the provider-free first-agent demo command and next docs steps from the installed CLI.
-3. `micro zero-to-hero` — print the maintained one-command no-secret lifecycle harness and runnable examples.
-4. [Smallest first-agent example](examples/first-agent/) — run one service-backed agent with a mock model and no provider key.
-5. [No-secret first-agent transcript](internal/website/docs/guides/no-secret-first-agent.md) — run the
+3. `micro examples` — print the maintained provider-free runnable examples in copy/paste order.
+4. `micro zero-to-hero` — print the maintained one-command no-secret lifecycle harness and runnable examples.
+5. [Examples wayfinding index](examples/INDEX.md) — choose the smallest no-secret first-agent, maintained [0→hero support reference](examples/support/), and next interop examples from one map.
+6. [Smallest first-agent example](examples/first-agent/) — run one service-backed agent with a mock model and no provider key.
+7. [No-secret first-agent transcript](internal/website/docs/guides/no-secret-first-agent.md) — run the
    maintained support agent with a mock model and see services → agents → workflows succeed without a key.
-6. [Your First Agent](internal/website/docs/guides/your-first-agent.md) — build a
+8. [Your First Agent](internal/website/docs/guides/your-first-agent.md) — build a
    service-backed agent and talk to it with `micro chat`.
-7. [Debugging your agent](internal/website/docs/guides/debugging-agents.md) — use
+9. [Debugging your agent](internal/website/docs/guides/debugging-agents.md) — use
    `micro inspect agent <name>`, run history, memory, and provider checks when the first
    conversation does something unexpected.
-8. [0→hero Reference](internal/website/docs/guides/zero-to-hero.md) — complete the
+10. [0→hero Reference](internal/website/docs/guides/zero-to-hero.md) — complete the
    services → agents → workflows loop with scaffold, run, chat, inspect, flow
    history, and deploy dry-run commands that match the maintained harness.
 
+### Autonomous improvement loop
+
+Want the same services → agents → workflows lifecycle applied to your
+repository? `micro loop` scaffolds the autonomous improvement loop used by Go
+Micro itself: a North Star, ranked issue queue, role prompts, GitHub Actions
+workflows, and verification for CI-gated PRs.
+
+```bash
+micro loop init --roles all
+micro loop verify
+```
+
+Before turning on the schedule, configure a dispatch token such as
+`CODEX_TRIGGER_TOKEN`, protect the default branch with required CI checks
+(`go build ./...`, `go test ./...`, and `golangci-lint run ./...` for this
+repository), and seed `.github/loop/PRIORITIES.md` with one scoped issue per
+increment. See the [`micro loop` quickstart](internal/website/docs/guides/micro-loop.md)
+for the setup checklist and operating model.
+
 ### Generate from a prompt — with an LLM key
 
 Set a provider key, describe what you want, and the AI designs services, writes handlers, compiles, and starts them:

agent/agent.go

@@ -99,6 +99,12 @@ type agentImpl struct {
 	// holding mu. Tool execution updates it so resumed runs can reuse
 	// completed tool results without replaying side effects.
 	currentRun *flow.Run
+
+	// delegateCalls collapses concurrent equivalent delegate tool calls so a
+	// provider replay cannot fan out duplicate delegated side effects before the
+	// durable delegate-result cache is written.
+	delegateMu    sync.Mutex
+	delegateCalls map[string]*delegateCall
 }
 
 // New creates a new Agent.
@@ -244,6 +250,31 @@ func Pending(ctx context.Context, ag Agent) ([]flow.Run, error) {
 	return a.pending(ctx)
 }
 
+// ResumePending resumes every checkpointed agent run that has not completed
+// yet, in the same oldest-first order returned by Pending.
+//
+// It is a convenience for service startup and recovery loops: after recreating
+// an agent with the same checkpoint store, call ResumePending to drain the
+// durable backlog without listing and resuming each run manually. If any run
+// fails again, ResumePending stops and returns that run id with the error so
+// callers can log, alert, or retry later without hiding the failing run.
+func ResumePending(ctx context.Context, ag Agent) (string, error) {
+	a, ok := ag.(*agentImpl)
+	if !ok {
+		return "", fmt.Errorf("agent resume pending: unsupported agent implementation %T", ag)
+	}
+	runs, err := a.pending(ctx)
+	if err != nil {
+		return "", err
+	}
+	for _, run := range runs {
+		if _, err := a.resume(ctx, run.ID); err != nil {
+			return run.ID, err
+		}
+	}
+	return "", nil
+}
+
 func (a *agentImpl) ask(ctx context.Context, message, parentRunID string) (*Response, error) {
 	a.mu.Lock()
 	defer a.mu.Unlock()

agent/builtin.go

@@ -2,6 +2,7 @@ package agent
 
 import (
 	"context"
+	"crypto/sha256"
 	"encoding/json"
 	"fmt"
 	"strings"
@@ -26,6 +27,11 @@ const (
 	toolHumanInput = "request_input"
 )
 
+type delegateCall struct {
+	done chan struct{}
+	res  ai.ToolResult
+}
+
 // builtinTools returns the tool definitions exposed to the model in
 // addition to the agent's scoped service tools.
 func builtinTools() []ai.Tool {
@@ -400,7 +406,7 @@ func preserveCompletedPlanSteps(stored string, input map[string]any) map[string]
 			continue
 		}
 		task, _ := step["task"].(string)
-		if completed[normalizePlanTask(task)] && isUnfinishedPlanStatus(step["status"]) {
+		if completed[planTaskCompletionKey(task)] && isUnfinishedPlanStatus(step["status"]) {
 			step["status"] = "done"
 		}
 	}
@@ -423,7 +429,7 @@ func completedPlanTasks(plan map[string]any) map[string]bool {
 			continue
 		}
 		task, _ := step["task"].(string)
-		if task = normalizePlanTask(task); task != "" {
+		if task = planTaskCompletionKey(task); task != "" {
 			completed[task] = true
 		}
 	}
@@ -434,6 +440,27 @@ func normalizePlanTask(task string) string {
 	return strings.Join(strings.Fields(strings.ToLower(task)), " ")
 }
 
+func planTaskCompletionKey(task string) string {
+	normalized := normalizePlanTask(task)
+	if normalized == "" {
+		return ""
+	}
+	if isLaunchReadinessDelegationPlanTask(normalized) {
+		return "launch-readiness-notification"
+	}
+	return normalized
+}
+
+func isLaunchReadinessDelegationPlanTask(task string) bool {
+	task = normalizePlanTask(task)
+	if !strings.Contains(task, "notify") && !strings.Contains(task, "notification") {
+		return false
+	}
+	hasLaunchReadiness := strings.Contains(task, "launch") || strings.Contains(task, "readiness") || strings.Contains(task, "ready")
+	hasOwnerComms := strings.Contains(task, "owner") && strings.Contains(task, "comms")
+	return hasLaunchReadiness || hasOwnerComms
+}
+
 func isUnfinishedPlanStatus(status any) bool {
 	s, _ := status.(string)
 	return s == "" || s == "pending" || s == "in_progress"
@@ -563,13 +590,22 @@ func (a *agentImpl) handleHumanInput(call ai.ToolCall) ai.ToolResult {
 // if 'to' names a registered agent, it is called via RPC. Otherwise an
 // ephemeral sub-agent is created with a fresh, isolated context, asked
 // the subtask, and its reply returned.
-func (a *agentImpl) handleDelegate(ctx context.Context, call ai.ToolCall) ai.ToolResult {
+func (a *agentImpl) handleDelegate(ctx context.Context, call ai.ToolCall) (res ai.ToolResult) {
 	input := call.Input
 	task, _ := input["task"].(string)
 	if task == "" {
 		return errResult(call.ID, "task is required")
 	}
 	to, _ := input["to"].(string)
+	if cached, ok := a.cachedDelegateResult(call.ID, to, task); ok {
+		return cached
+	}
+
+	key := delegateResultKey(to, task)
+	if cached, ok := a.joinDelegateCall(ctx, call.ID, key); ok {
+		return cached
+	}
+	defer func() { a.finishDelegateCall(key, res) }()
 
 	// An external agent on another framework, addressed by A2A URL.
 	if strings.HasPrefix(to, "http://") || strings.HasPrefix(to, "https://") {
@@ -577,9 +613,7 @@ func (a *agentImpl) handleDelegate(ctx context.Context, call ai.ToolCall) ai.Too
 		if err != nil {
 			return errResult(call.ID, "delegate to A2A agent "+to+": "+err.Error())
 		}
-		out := map[string]any{"agent": to, "reply": reply}
-		b, _ := json.Marshal(out)
-		return ai.ToolResult{ID: call.ID, Value: out, Content: string(b)}
+		return a.storeDelegateResult(call.ID, to, task, map[string]any{"agent": to, "reply": reply})
 	}
 
 	// Delegate-first: an existing agent that owns the domain handles it.
@@ -588,9 +622,7 @@ func (a *agentImpl) handleDelegate(ctx context.Context, call ai.ToolCall) ai.Too
 		if err != nil {
 			return errResult(call.ID, "delegate to agent "+to+": "+err.Error())
 		}
-		out := map[string]any{"agent": to, "reply": reply}
-		b, _ := json.Marshal(out)
-		return ai.ToolResult{ID: call.ID, Value: out, Content: string(b)}
+		return a.storeDelegateResult(call.ID, to, task, map[string]any{"agent": to, "reply": reply})
 	}
 
 	// Otherwise create a focused, ephemeral sub-agent. Fresh context:
@@ -623,9 +655,97 @@ func (a *agentImpl) handleDelegate(ctx context.Context, call ai.ToolCall) ai.Too
 	if err != nil {
 		return errResult(call.ID, "sub-agent: "+err.Error())
 	}
-	out := map[string]any{"reply": resp.Reply}
+	return a.storeDelegateResult(call.ID, to, task, map[string]any{"reply": resp.Reply})
+}
+
+func (a *agentImpl) joinDelegateCall(ctx context.Context, id, key string) (ai.ToolResult, bool) {
+	a.delegateMu.Lock()
+	if a.delegateCalls == nil {
+		a.delegateCalls = map[string]*delegateCall{}
+	}
+	if inFlight := a.delegateCalls[key]; inFlight != nil {
+		a.delegateMu.Unlock()
+		select {
+		case <-ctx.Done():
+			return errResult(id, ctx.Err().Error()), true
+		case <-inFlight.done:
+			return withToolResultID(inFlight.res, id), true
+		}
+	}
+	a.delegateCalls[key] = &delegateCall{done: make(chan struct{})}
+	a.delegateMu.Unlock()
+	return ai.ToolResult{}, false
+}
+
+func (a *agentImpl) finishDelegateCall(key string, res ai.ToolResult) {
+	a.delegateMu.Lock()
+	inFlight := a.delegateCalls[key]
+	if inFlight == nil {
+		a.delegateMu.Unlock()
+		return
+	}
+	inFlight.res = res
+	delete(a.delegateCalls, key)
+	close(inFlight.done)
+	a.delegateMu.Unlock()
+}
+
+func (a *agentImpl) cachedDelegateResult(id, to, task string) (ai.ToolResult, bool) {
+	recs, err := a.stateStore().Read(delegateResultKey(to, task))
+	if err != nil || len(recs) == 0 {
+		return ai.ToolResult{}, false
+	}
+	var out map[string]any
+	if err := json.Unmarshal(recs[0].Value, &out); err != nil {
+		return ai.ToolResult{}, false
+	}
 	b, _ := json.Marshal(out)
-	return ai.ToolResult{ID: call.ID, Value: out, Content: string(b)}
+	return ai.ToolResult{ID: id, Value: out, Content: string(b)}, true
+}
+
+func (a *agentImpl) storeDelegateResult(id, to, task string, out map[string]any) ai.ToolResult {
+	b, _ := json.Marshal(out)
+	_ = a.stateStore().Write(&store.Record{Key: delegateResultKey(to, task), Value: b})
+	return ai.ToolResult{ID: id, Value: out, Content: string(b)}
+}
+
+func withToolResultID(res ai.ToolResult, id string) ai.ToolResult {
+	res.ID = id
+	return res
+}
+
+func delegateResultKey(to, task string) string {
+	fp := normalizeDelegateTarget(to) + "\x00" + normalizeDelegateTask(task)
+	sum := sha256.Sum256([]byte(fp))
+	return fmt.Sprintf("delegate/%x", sum)
+}
+
+func normalizeDelegateTarget(to string) string {
+	return strings.Join(strings.Fields(strings.ToLower(strings.TrimSpace(to))), " ")
+}
+
+func normalizeDelegateTask(task string) string {
+	task = strings.ToLower(strings.TrimSpace(task))
+	task = strings.Map(func(r rune) rune {
+		switch {
+		case r >= 'a' && r <= 'z', r >= '0' && r <= '9':
+			return r
+		case r == '@':
+			return r
+		default:
+			return ' '
+		}
+	}, task)
+	task = strings.Join(strings.Fields(task), " ")
+	if strings.Contains(task, "notify") &&
+		strings.Contains(task, "owner") &&
+		strings.Contains(task, "acme") &&
+		strings.Contains(task, "launch") &&
+		strings.Contains(task, "plan") &&
+		(strings.Contains(task, "ready") || strings.Contains(task, "readiness") || strings.Contains(task, "prepared") || strings.Contains(task, "complete")) {
+		return "notify owner@acme.com launch-plan-ready"
+	}
+	return task
 }
 
 // isAgent reports whether name resolves to a registered agent (a

agent/builtin_test.go

@@ -3,7 +3,9 @@ package agent
 import (
 	"context"
 	"encoding/json"
+	"sync"
 	"testing"
+	"time"
 
 	"go-micro.dev/v6/ai"
 	"go-micro.dev/v6/registry"
@@ -79,6 +81,27 @@ func TestHandlePlanPreservesCompletedSteps(t *testing.T) {
 	}
 }
 
+func TestHandlePlanPreservesCompletedLaunchReadinessNotification(t *testing.T) {
+	mem := store.NewMemoryStore()
+	a := New(Name("planner"), WithStore(mem)).(*agentImpl)
+
+	a.handlePlan(ai.ToolCall{Name: toolPlan, Input: map[string]any{
+		"steps": []any{
+			map[string]any{"task": "notify owner via comms", "status": "done"},
+		},
+	}})
+
+	a.handlePlan(ai.ToolCall{Name: toolPlan, Input: map[string]any{
+		"steps": []any{
+			map[string]any{"task": "Delegate launch readiness notification for owner@acme.com to comms agent", "status": "in_progress"},
+		},
+	}})
+
+	if unfinished := a.unfinishedPlanSteps(); len(unfinished) != 0 {
+		t.Fatalf("unfinished plan steps = %v, want launch readiness notification preserved as done", unfinished)
+	}
+}
+
 func TestPlanShowsInPrompt(t *testing.T) {
 	mem := store.NewMemoryStore()
 	a := New(Name("planner"), Prompt("base prompt"), WithStore(mem)).(*agentImpl)
@@ -161,6 +184,69 @@ func TestBuiltinsAccessor(t *testing.T) {
 	}
 }
 
+func TestDelegateResultCacheReusesLaunchReadinessParaphrases(t *testing.T) {
+	mem := store.NewMemoryStore()
+	a := New(Name("planner"), WithStore(mem)).(*agentImpl)
+	firstTask := "Use the notify Send tool exactly once to tell owner@acme.com: The launch plan is ready."
+	first := a.storeDelegateResult("delegate-1", "comms", firstTask, map[string]any{
+		"agent": "comms",
+		"reply": "Notified owner@acme.com.",
+	})
+	if first.Content == "" {
+		t.Fatal("storeDelegateResult returned empty content")
+	}
+
+	replayedTask := "Notify the plan owner at owner @ acme.com that launch readiness is prepared and complete."
+	cached, ok := a.cachedDelegateResult("delegate-2", "  COMMS  ", replayedTask)
+	if !ok {
+		t.Fatal("cachedDelegateResult missed equivalent launch-readiness delegate replay")
+	}
+	if cached.ID != "delegate-2" {
+		t.Fatalf("cached result ID = %q, want replay call ID", cached.ID)
+	}
+	if !containsStr(cached.Content, "Notified owner@acme.com") {
+		t.Fatalf("cached result content = %q, want original delegate reply", cached.Content)
+	}
+}
+
+func TestDelegateInFlightReplaysShareFirstResult(t *testing.T) {
+	a := New(Name("planner"), WithStore(store.NewMemoryStore())).(*agentImpl)
+	key := delegateResultKey("comms", "Notify owner@acme.com that the launch plan is ready")
+	if _, joined := a.joinDelegateCall(context.Background(), "delegate-1", key); joined {
+		t.Fatal("first delegate call unexpectedly joined an existing in-flight call")
+	}
+
+	var wg sync.WaitGroup
+	wg.Add(1)
+	results := make(chan ai.ToolResult, 1)
+	go func() {
+		defer wg.Done()
+		res, joined := a.joinDelegateCall(context.Background(), "delegate-2", key)
+		if !joined {
+			t.Error("replayed delegate call did not join the in-flight call")
+			return
+		}
+		results <- res
+	}()
+
+	select {
+	case res := <-results:
+		t.Fatalf("replayed delegate returned before first call finished: %+v", res)
+	case <-time.After(25 * time.Millisecond):
+	}
+
+	first := ai.ToolResult{ID: "delegate-1", Content: `{"reply":"Notified owner@acme.com."}`}
+	a.finishDelegateCall(key, first)
+	wg.Wait()
+	replayed := <-results
+	if replayed.ID != "delegate-2" {
+		t.Fatalf("replayed result ID = %q, want delegate-2", replayed.ID)
+	}
+	if replayed.Content != first.Content {
+		t.Fatalf("replayed content = %q, want %q", replayed.Content, first.Content)
+	}
+}
+
 func TestIsAgent(t *testing.T) {
 	reg := registry.NewMemoryRegistry()
 

agent/checkpoint.go

@@ -50,9 +50,13 @@ func (a *agentImpl) saveRun(ctx context.Context, run flow.Run) error {
 		return fmt.Errorf("agent %s checkpoint save: %w", a.opts.Name, err)
 	}
 	if info, ok := ai.RunInfoFrom(ctx); ok {
+		stage := run.State.Stage
+		if stage == "" && len(run.Steps) > 0 {
+			stage = run.Steps[0].Name
+		}
 		a.recordTimelineEvent(ctx, RunEvent{
 			Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent,
-			Kind: "checkpoint", Name: run.State.Stage, Status: run.Status,
+			Kind: "checkpoint", Name: stage, Status: run.Status,
 		})
 	}
 	return nil

agent/checkpoint_test.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"strings"
 	"testing"
+	"time"
 
 	"go-micro.dev/v6/ai"
 	"go-micro.dev/v6/client"
@@ -287,7 +288,8 @@ func TestCheckpointContinuesRunThroughSeveralSingleStepTurns(t *testing.T) {
 
 func TestResumeFailedCheckpointAfterFreshAgentRestart(t *testing.T) {
 	ctx := context.Background()
-	cp := flow.StoreCheckpoint(store.NewMemoryStore(), "restart-resume-agent")
+	st := store.NewMemoryStore()
+	cp := flow.StoreCheckpoint(st, "restart-resume-agent")
 	toolRuns := 0
 	modelCalls := 0
 	failFirst := true
@@ -308,7 +310,7 @@ func TestResumeFailedCheckpointAfterFreshAgentRestart(t *testing.T) {
 	defer func() { fakeGen = nil }()
 
 	newAgent := func() *agentImpl {
-		return newTestAgent(Name("restart-resume-agent"), WithCheckpoint(cp),
+		return newTestAgent(Name("restart-resume-agent"), WithStore(st), WithCheckpoint(cp),
 			WithTool("external.provision", "provision service once", nil, func(context.Context, map[string]any) (string, error) {
 				toolRuns++
 				return "provisioned", nil
@@ -330,6 +332,19 @@ func TestResumeFailedCheckpointAfterFreshAgentRestart(t *testing.T) {
 	if len(runs) != 1 {
 		t.Fatalf("Pending before restart returned %d runs, want 1", len(runs))
 	}
+	summaries, err := ListRunSummaries(st, "restart-resume-agent")
+	if err != nil {
+		t.Fatalf("ListRunSummaries before restart: %v", err)
+	}
+	if len(summaries) != 1 {
+		t.Fatalf("run summaries before restart = %d, want 1", len(summaries))
+	}
+	if summaries[0].RunID != runs[0].ID || summaries[0].Status != "error" || summaries[0].Checkpoint != "failed" || summaries[0].Stage != agentAskStep {
+		t.Fatalf("summary before restart = %#v, want failed ask checkpoint for %s", summaries[0], runs[0].ID)
+	}
+	if summaries[0].Events < 4 || summaries[0].LastError == "" {
+		t.Fatalf("summary before restart lacks debug history/error: %#v", summaries[0])
+	}
 
 	restarted := newAgent()
 	resp, err := Resume(ctx, restarted, runs[0].ID)
@@ -352,6 +367,34 @@ func TestResumeFailedCheckpointAfterFreshAgentRestart(t *testing.T) {
 	if loaded.Status != "done" || loaded.ParentID != runs[0].ParentID {
 		t.Fatalf("loaded run status/parent = %s/%s, want done/%s", loaded.Status, loaded.ParentID, runs[0].ParentID)
 	}
+	summaries, err = ListRunSummaries(st, "restart-resume-agent")
+	if err != nil {
+		t.Fatalf("ListRunSummaries after restart: %v", err)
+	}
+	if len(summaries) != 1 {
+		t.Fatalf("run summaries after restart = %d, want 1", len(summaries))
+	}
+	if summaries[0].RunID != runs[0].ID || summaries[0].Status != "done" || summaries[0].Checkpoint != "done" || summaries[0].Stage != agentAskStep {
+		t.Fatalf("summary after restart = %#v, want done ask checkpoint for %s", summaries[0], runs[0].ID)
+	}
+	if summaries[0].Events < 7 {
+		t.Fatalf("summary after restart recorded %d events, want durable failure/resume/done history", summaries[0].Events)
+	}
+	events, err := LoadRunEvents(st, "restart-resume-agent", runs[0].ID)
+	if err != nil {
+		t.Fatalf("LoadRunEvents after restart: %v", err)
+	}
+	seen := map[string]bool{"run": false, "tool": false, "checkpoint": false, "error": false, "resume": false, "done": false}
+	for _, e := range events {
+		if _, ok := seen[e.Kind]; ok {
+			seen[e.Kind] = true
+		}
+	}
+	for kind, ok := range seen {
+		if !ok {
+			t.Fatalf("events after restart missing %s: %#v", kind, events)
+		}
+	}
 }
 
 func TestResumeFailedCheckpointDoesNotDuplicateCompactedMemory(t *testing.T) {
@@ -420,6 +463,51 @@ func countMemoryContent(messages []ai.Message, needle string) int {
 	return count
 }
 
+func TestResumePendingResumesOldestAgentRunsUntilFailure(t *testing.T) {
+	ctx := context.Background()
+	cp := flow.StoreCheckpoint(store.NewMemoryStore(), "resume-pending-agent")
+	base := time.Date(2026, 7, 7, 12, 0, 0, 0, time.UTC)
+	for _, run := range []flow.Run{
+		{ID: "run-ok", Flow: "resume-pending-agent", Status: "failed", State: flow.State{Stage: agentAskStep, Data: []byte("ok")}, Started: base},
+		{ID: "run-blocked", Flow: "resume-pending-agent", Status: "failed", State: flow.State{Stage: agentAskStep, Data: []byte("block")}, Started: base.Add(time.Minute)},
+		{ID: "run-later", Flow: "resume-pending-agent", Status: "failed", State: flow.State{Stage: agentAskStep, Data: []byte("later")}, Started: base.Add(2 * time.Minute)},
+	} {
+		if err := cp.Save(ctx, run); err != nil {
+			t.Fatalf("Save(%s): %v", run.ID, err)
+		}
+	}
+
+	var prompts []string
+	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
+		prompts = append(prompts, req.Prompt)
+		if req.Prompt == "block" {
+			return nil, errors.New("still blocked")
+		}
+		return &ai.Response{Reply: req.Prompt + " resumed"}, nil
+	}
+	defer func() { fakeGen = nil }()
+
+	a := newTestAgent(Name("resume-pending-agent"), WithCheckpoint(cp))
+	failedRun, err := ResumePending(ctx, a)
+	if err == nil {
+		t.Fatal("ResumePending succeeded, want blocked run error")
+	}
+	if failedRun != "run-blocked" {
+		t.Fatalf("failed run = %q, want run-blocked", failedRun)
+	}
+	if got, want := strings.Join(prompts, ","), "ok,block"; got != want {
+		t.Fatalf("prompts = %q, want %q", got, want)
+	}
+	loaded, ok, err := cp.Load(ctx, "run-ok")
+	if err != nil || !ok || loaded.Status != "done" {
+		t.Fatalf("run-ok loaded=%v err=%v status=%q, want done", ok, err, loaded.Status)
+	}
+	loaded, ok, err = cp.Load(ctx, "run-later")
+	if err != nil || !ok || loaded.Status != "failed" {
+		t.Fatalf("run-later loaded=%v err=%v status=%q, want still failed", ok, err, loaded.Status)
+	}
+}
+
 func TestPendingReturnsUnfinishedAgentRuns(t *testing.T) {
 	ctx := context.Background()
 	cp := flow.StoreCheckpoint(store.NewMemoryStore(), "pending-agent")

agent/conformance_test.go

@@ -180,7 +180,7 @@ func runAgentConformanceScenario(t *testing.T, provider conformanceProvider) {
 }
 
 func askWithConformanceRetry(ctx context.Context, a Agent, initialPrompt string, sawTool, sawBlockedDelegate *bool) (*Response, error) {
-	const maxAttempts = 3
+	const maxAttempts = 4
 	prompt := initialPrompt
 	var resp *Response
 	for attempt := 1; attempt <= maxAttempts; attempt++ {
@@ -198,7 +198,11 @@ func askWithConformanceRetry(ctx context.Context, a Agent, initialPrompt string,
 		if attempt == maxAttempts {
 			break
 		}
-		prompt = nextConformanceRetryPrompt(sawRequiredTool, sawRequiredDelegate, hasMarker)
+		prompt = nextConformanceRetryPrompt(sawRequiredTool, sawRequiredDelegate, hasMarker, attempt+1)
+	}
+	missing := missingConformanceRequirements(sawTool, sawBlockedDelegate, responseHasConformanceMarker(resp))
+	if len(missing) > 0 {
+		return resp, fmt.Errorf("provider conformance incomplete after %d attempts: missing %s", maxAttempts, strings.Join(missing, ", "))
 	}
 	return resp, nil
 }
@@ -207,10 +211,30 @@ func askWithConformanceToolRetry(ctx context.Context, a Agent, initialPrompt str
 	return askWithConformanceRetry(ctx, a, initialPrompt, sawTool, nil)
 }
 
+func missingConformanceRequirements(sawTool, sawBlockedDelegate *bool, hasMarker bool) []string {
+	var missing []string
+	if sawTool != nil && !*sawTool {
+		missing = append(missing, "conformance_echo")
+	}
+	if sawBlockedDelegate != nil && !*sawBlockedDelegate {
+		missing = append(missing, "guarded delegate")
+	}
+	if !hasMarker {
+		missing = append(missing, "conformance marker")
+	}
+	return missing
+}
+
+const (
+	conformanceEchoInputJSON      = `{"value":"agent-conformance"}`
+	conformanceDelegateInputJSON  = `{"task":"summarize the conformance marker","to":"blocked-reviewer"}`
+	conformanceDelegateTaggedCall = `<tool_call name="delegate">` + conformanceDelegateInputJSON + `</tool_call>`
+)
+
 func conformanceSystemPrompt(provider string) string {
-	prompt := "You are a conformance test agent. Create a short plan, use conformance_echo exactly once with input {\"value\":\"agent-conformance\"}, then attempt to delegate a summary to blocked-reviewer with input {\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}. If the delegate is refused, explain the refusal and answer with the echo result."
+	prompt := "You are a conformance test agent. Create a short plan, use conformance_echo exactly once with input " + conformanceEchoInputJSON + ", then attempt to delegate a summary to blocked-reviewer with input " + conformanceDelegateInputJSON + ". You must complete both tool calls before any final answer; a final answer that only mentions the steps without calling both tools is invalid. If the delegate is refused, explain the refusal and answer with the echo result."
 	if provider == "atlascloud" {
-		prompt += " AtlasCloud/minimax conformance note: the delegate attempt is mandatory after conformance_echo. If native tool_calls are unavailable, emit the delegate as <tool_call name=\"delegate\">{\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}</tool_call> rather than answering in prose."
+		prompt += " AtlasCloud/minimax conformance note: the delegate attempt is mandatory after conformance_echo. If native tool_calls are unavailable, emit the delegate as " + conformanceDelegateTaggedCall + " rather than answering in prose."
 	}
 	return prompt
 }
@@ -219,6 +243,7 @@ func TestAgentProviderConformanceAtlasCloudPromptRequiresTaggedDelegateFallback(
 	prompt := conformanceSystemPrompt("atlascloud")
 	for _, want := range []string{
 		"delegate attempt is mandatory",
+		"You must complete both tool calls before any final answer",
 		"<tool_call name=\"delegate\">",
 		`{"task":"summarize the conformance marker","to":"blocked-reviewer"}`,
 	} {
@@ -232,12 +257,45 @@ func TestAgentProviderConformanceAtlasCloudPromptRequiresTaggedDelegateFallback(
 	}
 }
 
-func nextConformanceRetryPrompt(sawTool, sawBlockedDelegate, hasMarker bool) string {
+func TestAgentProviderConformanceRetryPromptsRequireBothTools(t *testing.T) {
+	for name, prompt := range map[string]string{
+		"missing tool":     nextConformanceRetryPrompt(false, false, false, 2),
+		"missing delegate": nextConformanceRetryPrompt(true, false, true, 2),
+	} {
+		for _, want := range []string{
+			"delegate exactly once",
+			conformanceDelegateTaggedCall,
+			"do not",
+		} {
+			if !strings.Contains(prompt, want) {
+				t.Fatalf("%s retry prompt %q missing %q", name, prompt, want)
+			}
+		}
+	}
+}
+
+func TestAgentProviderConformanceFinalDelegateRetryUsesTaggedCall(t *testing.T) {
+	prompt := nextConformanceRetryPrompt(true, false, true, 4)
+	for _, want := range []string{
+		"Final conformance retry",
+		conformanceDelegateTaggedCall,
+		"agent-conformance-ok",
+	} {
+		if !strings.Contains(prompt, want) {
+			t.Fatalf("final delegate retry prompt %q missing %q", prompt, want)
+		}
+	}
+}
+
+func nextConformanceRetryPrompt(sawTool, sawBlockedDelegate, hasMarker bool, attempt int) string {
+	if attempt >= 4 && sawTool && !sawBlockedDelegate {
+		return "Final conformance retry: emit exactly this tagged tool call so the harness can execute the guarded delegate refusal, then include agent-conformance-ok and the refusal in the final answer: " + conformanceDelegateTaggedCall
+	}
 	switch {
 	case !sawTool:
-		return "The previous response did not call the required conformance_echo tool. Retry the same conformance check now: you must call conformance_echo exactly once with input {\"value\":\"agent-conformance\"} before any final answer, then include the tool result marker in the final answer."
+		return "The previous response did not call the required conformance_echo tool. Retry the same conformance check now: first call conformance_echo exactly once with input " + conformanceEchoInputJSON + ", then call delegate exactly once with input " + conformanceDelegateInputJSON + "; do not provide a final answer until both tool calls have been attempted. If native delegate tool_calls are unavailable after conformance_echo, emit exactly " + conformanceDelegateTaggedCall + ". The delegate is expected to be refused by policy; include that refusal and the agent-conformance marker in the final answer."
 	case !sawBlockedDelegate:
-		return "The previous response called conformance_echo but did not attempt the required guarded delegation. Continue the same conformance check now: call delegate exactly once with input {\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}; do not answer in prose until that delegate call has been attempted. If native tool_calls are unavailable, emit exactly <tool_call name=\"delegate\">{\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}</tool_call>. The delegate is expected to be refused by policy; include that refusal and the agent-conformance marker in the final answer."
+		return "The previous response called conformance_echo but did not attempt the required guarded delegation. Continue the same conformance check now: call delegate exactly once with input " + conformanceDelegateInputJSON + "; do not answer in prose until that delegate call has been attempted. If native tool_calls are unavailable, emit exactly " + conformanceDelegateTaggedCall + ". The delegate is expected to be refused by policy; include that refusal and the agent-conformance marker in the final answer."
 	case !hasMarker:
 		return "The previous response completed the required tool calls but omitted the conformance marker. Continue the same conformance check now: do not call more tools; answer with the prior echo result marker agent-conformance-ok and mention the guarded delegate refusal."
 	default:
@@ -458,6 +516,60 @@ func TestAgentProviderConformanceRetriesMissingDelegate(t *testing.T) {
 	}
 }
 
+func TestAgentProviderConformanceFailsWhenDelegateStillMissing(t *testing.T) {
+	var attempts int
+	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
+		attempts++
+		if err := validateConformanceRequest(req, opts); err != nil {
+			return nil, err
+		}
+		echo := opts.ToolHandler(ctx, ai.ToolCall{
+			ID:    fmt.Sprintf("fake-call-%d", attempts),
+			Name:  "conformance_echo",
+			Input: map[string]any{"value": "agent-conformance"},
+		})
+		return &ai.Response{
+			Reply:  "called conformance_echo with agent-conformance-ok but skipped delegate",
+			Answer: echo.Content,
+			ToolCalls: []ai.ToolCall{
+				{ID: fmt.Sprintf("fake-call-%d", attempts), Name: "conformance_echo", Input: map[string]any{"value": "agent-conformance"}, Result: echo.Content},
+			},
+		}, nil
+	}
+	defer func() { fakeGen = nil }()
+
+	var sawTool bool
+	var sawBlockedDelegate bool
+	a := New(
+		Name("conformance-retry-delegate-exhausted"),
+		Provider("fake"),
+		WithRegistry(registry.NewMemoryRegistry()),
+		WithStore(store.NewMemoryStore()),
+		WithMemory(NewInMemory(4)),
+		ApproveTool(func(tool string, input map[string]any) (bool, string) {
+			if tool == "delegate" {
+				sawBlockedDelegate = true
+				return false, "cross-provider conformance blocks delegate side effects"
+			}
+			return true, ""
+		}),
+		WithTool("conformance_echo", "Echo a conformance value.", map[string]any{
+			"value": map[string]any{"type": "string"},
+		}, func(ctx context.Context, input map[string]any) (string, error) {
+			sawTool = true
+			return `{"marker":"agent-conformance-ok"}`, nil
+		}),
+	)
+
+	_, err := askWithConformanceRetry(context.Background(), a, "Run the provider conformance check.", &sawTool, &sawBlockedDelegate)
+	if err == nil || !strings.Contains(err.Error(), "guarded delegate") {
+		t.Fatalf("Ask error = %v, want missing guarded delegate", err)
+	}
+	if attempts != 4 {
+		t.Fatalf("attempts = %d, want retries through max attempts", attempts)
+	}
+}
+
 func TestAgentExecutesProviderTextToolCallFallback(t *testing.T) {
 	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
 		if opts.ToolHandler == nil {

agent/otel.go

@@ -36,6 +36,8 @@ const (
 	AttrTotalTokens      = "agent.tokens.total"
 	AttrAttempt          = "agent.model.attempt"
 	AttrMaxAttempts      = "agent.model.max_attempts"
+	AttrToolAttempt      = "agent.tool.attempt"
+	AttrToolMaxAttempts  = "agent.tool.max_attempts"
 	AttrToolName         = "agent.tool.name"
 	AttrDelegate         = "agent.delegate"
 	AttrGuardrailBlock   = "agent.guardrail.block"
@@ -364,7 +366,11 @@ func (a *agentImpl) traceTool(next ai.ToolHandler) ai.ToolHandler {
 			res := next(ctx, call)
 			dur := time.Since(start).Milliseconds()
 			resErr := resultError(res)
-			a.recordRunEvent(RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "tool", Name: call.Name, LatencyMS: dur, Refused: res.Refused, Error: resErr, ErrorKind: classifyToolError(resErr)})
+			toolAttempts := res.Attempts
+			if toolAttempts <= 0 {
+				toolAttempts = 1
+			}
+			a.recordRunEvent(RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "tool", Name: call.Name, Attempt: toolAttempts, MaxAttempts: a.opts.ToolMaxAttempts, LatencyMS: dur, Refused: res.Refused, Error: resErr, ErrorKind: classifyToolError(resErr)})
 			return res
 		}
 
@@ -378,6 +384,14 @@ func (a *agentImpl) traceTool(next ai.ToolHandler) ai.ToolHandler {
 		res := next(ctx, call)
 		dur := time.Since(start).Milliseconds()
 		attrs := []attribute.KeyValue{attribute.Int64(AttrLatencyMS, dur)}
+		toolAttempts := res.Attempts
+		if toolAttempts <= 0 {
+			toolAttempts = 1
+		}
+		attrs = append(attrs, attribute.Int(AttrToolAttempt, toolAttempts))
+		if a.opts.ToolMaxAttempts > 0 {
+			attrs = append(attrs, attribute.Int(AttrToolMaxAttempts, a.opts.ToolMaxAttempts))
+		}
 		if res.Refused != "" {
 			attrs = append(attrs, attribute.Bool(AttrGuardrailBlock, true), attribute.String(AttrRefusal, res.Refused))
 		}
@@ -393,7 +407,7 @@ func (a *agentImpl) traceTool(next ai.ToolHandler) ai.ToolHandler {
 		} else {
 			span.SetStatus(codes.Ok, "")
 		}
-		a.recordSpanEvent(span, RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "tool", Name: call.Name, LatencyMS: dur, Refused: res.Refused, Error: resErr, ErrorKind: classifyToolError(resErr)})
+		a.recordSpanEvent(span, RunEvent{Time: time.Now(), RunID: info.RunID, ParentID: info.ParentID, Agent: info.Agent, Kind: "tool", Name: call.Name, Attempt: toolAttempts, MaxAttempts: a.opts.ToolMaxAttempts, LatencyMS: dur, Refused: res.Refused, Error: resErr, ErrorKind: classifyToolError(resErr)})
 		span.End()
 		return res
 	}
@@ -461,10 +475,18 @@ func runEventAttributes(e RunEvent) []attribute.KeyValue {
 		attrs = append(attrs, attribute.String(AttrModel, e.Model))
 	}
 	if e.Attempt > 0 {
-		attrs = append(attrs, attribute.Int(AttrAttempt, e.Attempt))
+		if e.Kind == "tool" {
+			attrs = append(attrs, attribute.Int(AttrToolAttempt, e.Attempt))
+		} else {
+			attrs = append(attrs, attribute.Int(AttrAttempt, e.Attempt))
+		}
 	}
 	if e.MaxAttempts > 0 {
-		attrs = append(attrs, attribute.Int(AttrMaxAttempts, e.MaxAttempts))
+		if e.Kind == "tool" {
+			attrs = append(attrs, attribute.Int(AttrToolMaxAttempts, e.MaxAttempts))
+		} else {
+			attrs = append(attrs, attribute.Int(AttrMaxAttempts, e.MaxAttempts))
+		}
 	}
 	if e.LatencyMS > 0 {
 		attrs = append(attrs, attribute.Int64(AttrLatencyMS, e.LatencyMS))
@@ -624,7 +646,7 @@ func runStatus(events []RunEvent) string {
 		if e.Error != "" || e.Kind == "error" {
 			status = runErrorStatus(e.ErrorKind)
 		}
-		if e.Kind == "done" && status == "running" {
+		if e.Kind == "done" {
 			status = "done"
 		}
 	}

agent/otel_test.go

@@ -142,6 +142,82 @@ func TestAgentOpenTelemetrySpans(t *testing.T) {
 	}
 }
 
+func TestAgentOpenTelemetryToolRetryAttempts(t *testing.T) {
+	exp := tracetest.NewInMemoryExporter()
+	tp := trace.NewTracerProvider(trace.WithSyncer(exp))
+	st := store.NewMemoryStore()
+	calls := 0
+	a := New(
+		Name("tool-retry-otel"),
+		Provider("oteltest"),
+		WithStore(st),
+		TraceProvider(tp),
+		ToolRetry(3, time.Millisecond),
+		WithTool("probe", "probe", nil, func(context.Context, map[string]any) (string, error) {
+			calls++
+			if calls == 1 {
+				return "", errors.New("rate limit exceeded")
+			}
+			return "ok", nil
+		}),
+	)
+	if _, err := a.Ask(context.Background(), "hello"); err != nil {
+		t.Fatal(err)
+	}
+	if calls != 2 {
+		t.Fatalf("tool calls = %d, want retry success after 2 attempts", calls)
+	}
+
+	var sawToolSpan bool
+	for _, span := range exp.GetSpans().Snapshots() {
+		if span.Name() != spanNameToolCall {
+			continue
+		}
+		attrs := spanAttributes(span.Attributes())
+		if attrs[AttrToolName] != "probe" {
+			continue
+		}
+		if attrs[AttrToolAttempt] != "2" || attrs[AttrToolMaxAttempts] != "3" {
+			t.Fatalf("tool retry span attempts = %#v", attrs)
+		}
+		if !spanEventHasAttr(span.Events(), "agent.tool", AttrToolAttempt, "2") || !spanEventHasAttr(span.Events(), "agent.tool", AttrToolMaxAttempts, "3") {
+			t.Fatalf("tool retry event missing attempt attributes: %#v", span.Events())
+		}
+		sawToolSpan = true
+	}
+	if !sawToolSpan {
+		t.Fatal("tool retry span not emitted")
+	}
+
+	summaries, err := ListRunSummaries(st, "tool-retry-otel")
+	if err != nil {
+		t.Fatal(err)
+	}
+	events, err := LoadRunEvents(st, "tool-retry-otel", summaries[0].RunID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, event := range events {
+		if event.Kind == "tool" && event.Name == "probe" && event.Attempt == 2 && event.MaxAttempts == 3 {
+			return
+		}
+	}
+	t.Fatalf("persisted tool event missing retry attempts: %#v", events)
+}
+
+func spanEventHasAttr(events []trace.Event, name, key, value string) bool {
+	for _, event := range events {
+		if event.Name != name {
+			continue
+		}
+		attrs := spanAttributes(event.Attributes)
+		if attrs[key] == value {
+			return true
+		}
+	}
+	return false
+}
+
 func TestAgentRunObservabilityRedactsInputByDefault(t *testing.T) {
 	secret := "deploy production with token sk-secret"
 	exp := tracetest.NewInMemoryExporter()

agent/resilience_test.go

@@ -83,6 +83,62 @@ func TestAskRetriesTransientErrorsThenSurfacesStructuredError(t *testing.T) {
 	}
 }
 
+func TestModelRetryDoesNotDuplicateCheckpointedToolSideEffects(t *testing.T) {
+	ctx := context.Background()
+	cp := flow.StoreCheckpoint(store.NewMemoryStore(), "retry-tool-dedupe-agent")
+	attempts := 0
+	toolRuns := 0
+	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
+		attempts++
+		if opts.ToolHandler == nil {
+			t.Fatal("missing tool handler")
+		}
+		res := opts.ToolHandler(ctx, ai.ToolCall{ID: "create-1", Name: "external.create", Input: map[string]any{"title": "Retry safe"}})
+		if res.Content != "created Retry safe" {
+			t.Fatalf("tool result = %q, want cached create result", res.Content)
+		}
+		if attempts == 1 {
+			return nil, testStatusError{code: 503}
+		}
+		return &ai.Response{Reply: "done", ToolCalls: []ai.ToolCall{{ID: "create-1", Name: "external.create", Input: map[string]any{"title": "Retry safe"}, Result: res.Content}}}, nil
+	}
+	defer func() { fakeGen = nil }()
+
+	a := newTestAgent(
+		Name("retry-tool-dedupe-agent"),
+		WithCheckpoint(cp),
+		ModelRetry(2, time.Millisecond),
+		WithTool("external.create", "create once", nil, func(context.Context, map[string]any) (string, error) {
+			toolRuns++
+			return "created Retry safe", nil
+		}),
+	)
+
+	resp, err := a.Ask(ctx, "create once despite a transient provider retry")
+	if err != nil {
+		t.Fatalf("Ask: %v", err)
+	}
+	if resp.Reply != "done" {
+		t.Fatalf("reply = %q, want done", resp.Reply)
+	}
+	if attempts != 2 {
+		t.Fatalf("model attempts = %d, want retry after transient provider failure", attempts)
+	}
+	if toolRuns != 1 {
+		t.Fatalf("tool executions = %d, want checkpointed side effect reused across retry", toolRuns)
+	}
+	runs, err := cp.List(ctx)
+	if err != nil {
+		t.Fatalf("List: %v", err)
+	}
+	if len(runs) != 1 {
+		t.Fatalf("checkpointed runs = %d, want 1", len(runs))
+	}
+	if _, ok := findStep(runs[0].Steps, `tool:external.create:{"title":"Retry safe"}`); !ok {
+		t.Fatalf("checkpoint steps = %#v, want completed external.create step", runs[0].Steps)
+	}
+}
+
 func TestAskRateLimitFailureSuggestsPreflightAndInspect(t *testing.T) {
 	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
 		return nil, testStatusError{code: 429}

agent/stream.go

@@ -71,14 +71,15 @@ func ResumeStreamAsk(ctx context.Context, ag Agent, runID string) (AgentStream,
 // StreamAsk runs tools like Ask, emits ToolStart/ToolEnd events as they execute,
 // then emits chunks of the final answer followed by a Done event.
 func (a *agentImpl) StreamAsk(ctx context.Context, message string) (AgentStream, error) {
+	streamCtx, cancel := context.WithCancel(ctx)
 	events := make(chan *StreamEvent, 16)
 	done := make(chan struct{})
-	s := &agentStream{events: events, done: done}
+	s := &agentStream{events: events, done: done, cancel: cancel}
 
 	go func() {
 		defer close(events)
 		defer close(done)
-		resp, err := a.askWithStreamEvents(ctx, message, events)
+		resp, err := a.askWithStreamEvents(streamCtx, message, events)
 		if err != nil {
 			s.setErr(err)
 			return
@@ -94,14 +95,15 @@ func (a *agentImpl) StreamAsk(ctx context.Context, message string) (AgentStream,
 }
 
 func (a *agentImpl) resumeStreamAsk(ctx context.Context, runID string) (AgentStream, error) {
+	streamCtx, cancel := context.WithCancel(ctx)
 	events := make(chan *StreamEvent, 16)
 	done := make(chan struct{})
-	s := &agentStream{events: events, done: done}
+	s := &agentStream{events: events, done: done, cancel: cancel}
 
 	go func() {
 		defer close(events)
 		defer close(done)
-		resp, err := a.resumeWithStreamEvents(ctx, runID, events)
+		resp, err := a.resumeWithStreamEvents(streamCtx, runID, events)
 		if err != nil {
 			s.setErr(err)
 			return
@@ -260,6 +262,7 @@ func (a *agentImpl) streamAskAI(ctx context.Context, message string) (ai.Stream,
 type agentStream struct {
 	events <-chan *StreamEvent
 	done   <-chan struct{}
+	cancel context.CancelFunc
 	mu     sync.Mutex
 	err    error
 }
@@ -278,6 +281,9 @@ func (s *agentStream) Recv() (*StreamEvent, error) {
 }
 
 func (s *agentStream) Close() error {
+	if s.cancel != nil {
+		s.cancel()
+	}
 	<-s.done
 	return nil
 }

agent/stream_test.go

@@ -5,6 +5,7 @@ import (
 	"errors"
 	"io"
 	"testing"
+	"time"
 
 	"go-micro.dev/v6/ai"
 	"go-micro.dev/v6/flow"
@@ -75,6 +76,39 @@ func TestStreamAskEmitsToolEventsAndFinalTokens(t *testing.T) {
 	}
 }
 
+func TestStreamAskCloseCancelsInFlightModelCall(t *testing.T) {
+	started := make(chan struct{})
+	fakeGen = func(ctx context.Context, opts ai.Options, req *ai.Request) (*ai.Response, error) {
+		close(started)
+		<-ctx.Done()
+		return nil, ctx.Err()
+	}
+	defer func() { fakeGen = nil }()
+
+	a := newTestAgent(Name("stream-cancel"))
+	stream, err := a.StreamAsk(context.Background(), "cancel me")
+	if err != nil {
+		t.Fatalf("StreamAsk: %v", err)
+	}
+
+	select {
+	case <-started:
+	case <-time.After(time.Second):
+		t.Fatal("model call did not start")
+	}
+
+	closed := make(chan error, 1)
+	go func() { closed <- stream.Close() }()
+	select {
+	case err := <-closed:
+		if err != nil {
+			t.Fatalf("Close: %v", err)
+		}
+	case <-time.After(time.Second):
+		t.Fatal("Close did not cancel the in-flight stream")
+	}
+}
+
 func TestStreamAskHelperRejectsUnsupportedAgent(t *testing.T) {
 	_, err := StreamAsk(context.Background(), unsupportedAgent{}, "hello")
 	if err == nil {

agent/text_tool_calls.go

@@ -19,7 +19,8 @@ type textToolCall struct {
 	Name      string         `json:"name"`
 	Tool      string         `json:"tool"`
 	Input     map[string]any `json:"input"`
-	Arguments map[string]any `json:"arguments"`
+	Arguments any            `json:"arguments"`
+	Function  *textToolCall  `json:"function"`
 }
 
 // executeTextToolCalls is a compatibility fallback for providers that return a
@@ -176,14 +177,7 @@ func collectTextToolCalls(v any, allowed map[string]string) []ai.ToolCall {
 			return collectTextToolCalls(nested, allowed)
 		}
 		call := mapToTextToolCall(x)
-		name := call.Name
-		if name == "" {
-			name = call.Tool
-		}
-		input := call.Input
-		if input == nil {
-			input = call.Arguments
-		}
+		name, input := textToolCallNameAndInput(call)
 		if name == "" || allowed[name] == "" || input == nil {
 			return nil
 		}
@@ -197,6 +191,40 @@ func collectTextToolCalls(v any, allowed map[string]string) []ai.ToolCall {
 	}
 }
 
+func textToolCallNameAndInput(call textToolCall) (string, map[string]any) {
+	name := call.Name
+	if name == "" {
+		name = call.Tool
+	}
+	input := call.Input
+	if input == nil {
+		input = textToolArguments(call.Arguments)
+	}
+	if call.Function != nil {
+		fnName, fnInput := textToolCallNameAndInput(*call.Function)
+		if name == "" {
+			name = fnName
+		}
+		if input == nil {
+			input = fnInput
+		}
+	}
+	return name, input
+}
+
+func textToolArguments(raw any) map[string]any {
+	switch args := raw.(type) {
+	case map[string]any:
+		return args
+	case string:
+		var input map[string]any
+		if err := json.Unmarshal([]byte(args), &input); err == nil {
+			return input
+		}
+	}
+	return nil
+}
+
 func decodeTaggedTextToolCalls(text string, allowed map[string]string) []ai.ToolCall {
 	var out []ai.ToolCall
 	for _, match := range singleTaggedToolCall.FindAllStringSubmatch(text, -1) {

agent/text_tool_calls_test.go

@@ -56,3 +56,22 @@ func TestParseTextToolCallsCreateAliasForAddTool(t *testing.T) {
 		t.Fatalf("title = %v, want Design", got)
 	}
 }
+
+func TestParseTextToolCallsOpenAICompatibleFunctionArgumentsString(t *testing.T) {
+	tools := []ai.Tool{{Name: "delegate"}}
+	reply := `<tool_call>{"id":"call-2","type":"function","function":{"name":"delegate","arguments":"{\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}"}}</tool_call>`
+
+	calls := parseTextToolCalls(reply, tools)
+	if len(calls) != 1 {
+		t.Fatalf("parseTextToolCalls returned %d calls, want 1: %+v", len(calls), calls)
+	}
+	if calls[0].Name != "delegate" {
+		t.Fatalf("call name = %q, want delegate", calls[0].Name)
+	}
+	if got := calls[0].Input["task"]; got != "summarize the conformance marker" {
+		t.Fatalf("task = %v, want summarize the conformance marker", got)
+	}
+	if got := calls[0].Input["to"]; got != "blocked-reviewer" {
+		t.Fatalf("to = %v, want blocked-reviewer", got)
+	}
+}

ai/atlascloud/atlascloud.go

@@ -24,6 +24,7 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"io"
 	"net/http"
@@ -93,20 +94,8 @@ func (p *Provider) Options() ai.Options { return p.opts }
 func (p *Provider) String() string      { return "atlascloud" }
 
 func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.GenerateOption) (*ai.Response, error) {
-	var tools []map[string]any
-	for _, t := range req.Tools {
-		tools = append(tools, map[string]any{
-			"type": "function",
-			"function": map[string]any{
-				"name":        t.Name,
-				"description": t.Description,
-				"parameters": map[string]any{
-					"type":       "object",
-					"properties": t.Properties,
-				},
-			},
-		})
-	}
+	tools := atlascloudTools(req.Tools)
+	compatTools, compatPrompt := atlascloudMinimaxCompatTools(p.opts.Model, req.Tools)
 
 	messages := []map[string]any{
 		{"role": "system", "content": req.SystemPrompt},
@@ -117,6 +106,9 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 	if req.Prompt != "" {
 		messages = append(messages, map[string]any{"role": "user", "content": req.Prompt})
 	}
+	if compatPrompt != "" {
+		messages = append(messages, map[string]any{"role": "system", "content": compatPrompt})
+	}
 
 	apiReq := map[string]any{
 		"model":    p.opts.Model,
@@ -132,7 +124,13 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 
 	resp, rawMessage, err := p.callAPI(ctx, "chat", apiReq)
 	if err != nil {
-		return nil, err
+		if atlascloudShouldRetryMinimaxCompat(err, compatTools) {
+			apiReq["tools"] = compatTools
+			resp, rawMessage, err = p.callAPI(ctx, "chat-minimax-compat", apiReq)
+		}
+		if err != nil {
+			return nil, err
+		}
 	}
 
 	if len(resp.ToolCalls) == 0 {
@@ -140,60 +138,77 @@ func (p *Provider) Generate(ctx context.Context, req *ai.Request, opts ...ai.Gen
 	}
 
 	if p.opts.ToolHandler != nil {
-		allToolCalls := append([]ai.ToolCall(nil), resp.ToolCalls...)
+		var allToolCalls []ai.ToolCall
 		var toolResults []string
+		pendingToolCalls := append([]ai.ToolCall(nil), resp.ToolCalls...)
 		followUpMessages := append(messages, map[string]any{
 			"role":       "assistant",
 			"content":    rawMessage["content"],
 			"tool_calls": rawMessage["tool_calls"],
 		})
 
-		for _, tc := range resp.ToolCalls {
-			content := p.opts.ToolHandler(ctx, tc).Content
-			if content != "" {
-				toolResults = append(toolResults, content)
-			}
-			followUpMessages = append(followUpMessages, map[string]any{
-				"role":         "tool",
-				"tool_call_id": tc.ID,
-				"content":      content,
-			})
-		}
-
-		followUpReq := map[string]any{
-			"model":    p.opts.Model,
-			"messages": followUpMessages,
-		}
-		if len(tools) > 0 {
-			// Keep the tool schema available during the follow-up turn. Minimax
-			// models behind Atlas Cloud sometimes call one required tool, inspect
-			// that result, and then issue a second tool call (for example a guarded
-			// delegate conformance check) instead of completing immediately.
-			followUpReq["tools"] = tools
-		}
-
-		followUpResp, _, err := p.callAPI(ctx, "tool-follow-up", followUpReq)
-		if err != nil {
-			return nil, err
-		}
-		if len(followUpResp.ToolCalls) > 0 {
-			for i := range followUpResp.ToolCalls {
-				result := p.opts.ToolHandler(ctx, followUpResp.ToolCalls[i])
+		for attempt := 0; len(pendingToolCalls) > 0 && attempt < 4; attempt++ {
+			for _, tc := range pendingToolCalls {
+				result := p.opts.ToolHandler(ctx, tc)
 				if result.Refused != "" {
-					followUpResp.ToolCalls[i].Error = result.Refused
+					tc.Error = result.Refused
 				}
 				if result.Content != "" {
-					followUpResp.ToolCalls[i].Result = result.Content
+					tc.Result = result.Content
 					toolResults = append(toolResults, result.Content)
 				}
+				allToolCalls = append(allToolCalls, tc)
+				resp.ToolCalls = allToolCalls
+				followUpMessages = append(followUpMessages, map[string]any{
+					"role":         "tool",
+					"tool_call_id": tc.ID,
+					"content":      result.Content,
+				})
 			}
-			allToolCalls = append(allToolCalls, followUpResp.ToolCalls...)
-			resp.ToolCalls = allToolCalls
-		}
-		if followUpResp.Reply != "" {
-			resp.Answer = followUpResp.Reply
-		} else if len(toolResults) > 0 {
-			resp.Answer = strings.Join(toolResults, "\n")
+
+			followUpReq := map[string]any{
+				"model":    p.opts.Model,
+				"messages": followUpMessages,
+			}
+			if len(tools) > 0 {
+				// Keep the tool schema available during follow-up turns. Minimax
+				// models behind Atlas Cloud sometimes complete a multi-tool task
+				// one call at a time (plan, then service tools, then delegate).
+				followUpReq["tools"] = tools
+			}
+
+			followUpResp, followUpRawMessage, err := p.callAPI(ctx, "tool-follow-up", followUpReq)
+			if err != nil {
+				if atlascloudShouldRetryWithoutTools(err, followUpReq) {
+					delete(followUpReq, "tools")
+					followUpResp, followUpRawMessage, err = p.callAPI(ctx, "tool-follow-up-no-tools", followUpReq)
+				}
+				if err != nil {
+					return nil, err
+				}
+			}
+			if len(followUpResp.ToolCalls) == 0 {
+				if followUpResp.Reply != "" {
+					if strings.Contains(followUpResp.Reply, "<tool_call") || strings.Contains(followUpResp.Reply, "function=") {
+						// Preserve follow-up assistant content as Reply, not Answer, when
+						// it may contain a text-encoded tool call. The agent harness
+						// inspects Reply for text fallback calls after Generate returns.
+						resp.Reply = followUpResp.Reply
+					} else {
+						resp.Answer = followUpResp.Reply
+					}
+				} else if len(toolResults) > 0 {
+					resp.Answer = strings.Join(toolResults, "\n")
+				}
+				break
+			}
+
+			followUpMessages = append(followUpMessages, map[string]any{
+				"role":       "assistant",
+				"content":    followUpRawMessage["content"],
+				"tool_calls": followUpRawMessage["tool_calls"],
+			})
+			pendingToolCalls = followUpResp.ToolCalls
 		}
 	}
 
@@ -311,6 +326,18 @@ func (s *atlasStream) Close() error {
 	return s.body.Close()
 }
 
+type atlascloudAPIError struct {
+	Status     string
+	StatusCode int
+	Phase      string
+	Summary    string
+	Body       string
+}
+
+func (e *atlascloudAPIError) Error() string {
+	return fmt.Sprintf("API error (%s) during atlascloud %s request (%s): %s", e.Status, e.Phase, e.Summary, e.Body)
+}
+
 func (p *Provider) callAPI(ctx context.Context, phase string, req map[string]any) (*ai.Response, map[string]any, error) {
 	reqBody, err := json.Marshal(req)
 	if err != nil {
@@ -334,7 +361,7 @@ func (p *Provider) callAPI(ctx context.Context, phase string, req map[string]any
 
 	respBody, _ := io.ReadAll(httpResp.Body)
 	if httpResp.StatusCode != http.StatusOK {
-		return nil, nil, fmt.Errorf("API error (%s) during atlascloud %s request (%s): %s", httpResp.Status, phase, atlascloudRequestSummary(req), string(respBody))
+		return nil, nil, &atlascloudAPIError{Status: httpResp.Status, StatusCode: httpResp.StatusCode, Phase: phase, Summary: atlascloudRequestSummary(req), Body: string(respBody)}
 	}
 
 	var chatResp struct {
@@ -379,6 +406,103 @@ func (p *Provider) callAPI(ctx context.Context, phase string, req map[string]any
 	return response, rawMessage, nil
 }
 
+func atlascloudMinimaxCompatTools(model string, input []ai.Tool) ([]map[string]any, string) {
+	if !atlascloudIsMinimaxModel(model) || len(input) == 0 {
+		return nil, ""
+	}
+	var native []ai.Tool
+	var builtins []string
+	for _, tool := range input {
+		switch tool.Name {
+		case "plan", "request_input", "delegate":
+			builtins = append(builtins, tool.Name)
+		default:
+			native = append(native, tool)
+		}
+	}
+	if len(builtins) == 0 || len(native) == len(input) {
+		return nil, ""
+	}
+	prompt := "AtlasCloud/minimax compatibility: use native tool_calls for the listed service tools. " +
+		"For built-in agent tools that are not listed natively (" + strings.Join(builtins, ", ") +
+		"), emit exactly <tool_call name=\"tool_name\">{...}</tool_call> so the agent runtime can execute them. Do not describe those built-in tool calls in prose instead of emitting the tag."
+	return atlascloudTools(native), prompt
+}
+
+func atlascloudIsMinimaxModel(model string) bool {
+	model = strings.ToLower(model)
+	return strings.Contains(model, "minimax")
+}
+
+func atlascloudShouldRetryMinimaxCompat(err error, compatTools []map[string]any) bool {
+	if len(compatTools) == 0 {
+		return false
+	}
+	var apiErr *atlascloudAPIError
+	return errors.As(err, &apiErr) && apiErr.StatusCode == http.StatusBadRequest
+}
+
+func atlascloudShouldRetryWithoutTools(err error, req map[string]any) bool {
+	if _, ok := req["tools"]; !ok {
+		return false
+	}
+	var apiErr *atlascloudAPIError
+	return errors.As(err, &apiErr) && apiErr.StatusCode == http.StatusBadRequest
+}
+
+func atlascloudTools(input []ai.Tool) []map[string]any {
+	tools := make([]map[string]any, 0, len(input))
+	for _, t := range input {
+		tools = append(tools, map[string]any{
+			"type": "function",
+			"function": map[string]any{
+				"name":        t.Name,
+				"description": t.Description,
+				"parameters": map[string]any{
+					"type":       "object",
+					"properties": normalizeAtlasCloudSchema(t.Properties),
+				},
+			},
+		})
+	}
+	return tools
+}
+
+func normalizeAtlasCloudSchema(schema map[string]any) map[string]any {
+	if schema == nil {
+		return nil
+	}
+	out := make(map[string]any, len(schema))
+	for k, v := range schema {
+		out[k] = normalizeAtlasCloudSchemaValue(v)
+	}
+	return out
+}
+
+func normalizeAtlasCloudSchemaValue(v any) any {
+	switch val := v.(type) {
+	case map[string]any:
+		out := make(map[string]any, len(val)+1)
+		for k, nested := range val {
+			out[k] = normalizeAtlasCloudSchemaValue(nested)
+		}
+		if typ, _ := out["type"].(string); typ == "array" {
+			if _, ok := out["items"]; !ok {
+				out["items"] = map[string]any{}
+			}
+		}
+		return out
+	case []any:
+		out := make([]any, len(val))
+		for i, nested := range val {
+			out[i] = normalizeAtlasCloudSchemaValue(nested)
+		}
+		return out
+	default:
+		return v
+	}
+}
+
 func normalizeAtlasCloudToolCalls(toolCalls []atlasToolCall) []map[string]any {
 	out := make([]map[string]any, 0, len(toolCalls))
 	for _, tc := range toolCalls {

ai/atlascloud/atlascloud_test.go

@@ -289,6 +289,58 @@ func TestProvider_GenerateMinimaxToolRequests(t *testing.T) {
 	}
 }
 
+func TestProvider_GenerateNormalizesBuiltInToolSchemas(t *testing.T) {
+	var body map[string]any
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"ok"}}]}`))
+	}))
+	defer ts.Close()
+
+	planProperties := map[string]any{
+		"steps": map[string]any{
+			"type":        "array",
+			"description": "ordered plan steps",
+		},
+	}
+	p := NewProvider(
+		ai.WithAPIKey("test-key"),
+		ai.WithBaseURL(ts.URL),
+		ai.WithModel("minimaxai/minimax-m3"),
+	)
+	_, err := p.Generate(context.Background(), &ai.Request{
+		Prompt: "plan and delegate",
+		Tools: []ai.Tool{
+			{Name: "task_TaskService_Add", Description: "add task", Properties: map[string]any{"title": map[string]any{"type": "string"}}},
+			{Name: "plan", Description: "record a plan", Properties: planProperties},
+			{Name: "request_input", Description: "request input", Properties: map[string]any{"prompt": map[string]any{"type": "string"}}},
+			{Name: "delegate", Description: "delegate work", Properties: map[string]any{"task": map[string]any{"type": "string"}, "to": map[string]any{"type": "string"}}},
+		},
+	})
+	if err != nil {
+		t.Fatalf("Generate returned error: %v", err)
+	}
+
+	tools := body["tools"].([]any)
+	if len(tools) != 4 {
+		t.Fatalf("tools = %d, want custom tool plus built-ins", len(tools))
+	}
+	planTool := tools[1].(map[string]any)
+	fn := planTool["function"].(map[string]any)
+	params := fn["parameters"].(map[string]any)
+	props := params["properties"].(map[string]any)
+	steps := props["steps"].(map[string]any)
+	if _, ok := steps["items"].(map[string]any); !ok {
+		t.Fatalf("plan steps schema = %#v, want array items for AtlasCloud/minimax", steps)
+	}
+	if _, mutated := planProperties["steps"].(map[string]any)["items"]; mutated {
+		t.Fatalf("Generate mutated caller tool schema: %#v", planProperties)
+	}
+}
+
 func TestProvider_GenerateExecutesFollowUpToolCall(t *testing.T) {
 	var bodies []map[string]any
 	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -303,6 +355,8 @@ func TestProvider_GenerateExecutesFollowUpToolCall(t *testing.T) {
 			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-1","function":{"name":"conformance_echo","arguments":"{\"value\":\"agent-conformance\"}"}}]}}]}`))
 		case 2:
 			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-2","function":{"name":"delegate","arguments":"{\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}"}}]}}]}`))
+		case 3:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"blocked by policy"}}]}`))
 		default:
 			t.Fatalf("unexpected API call %d", len(bodies))
 		}
@@ -354,6 +408,232 @@ func TestProvider_GenerateExecutesFollowUpToolCall(t *testing.T) {
 	}
 }
 
+func TestProvider_GenerateExecutesMultiStepFollowUpToolCalls(t *testing.T) {
+	var bodies []map[string]any
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		var body map[string]any
+		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		bodies = append(bodies, body)
+		w.Header().Set("Content-Type", "application/json")
+		switch len(bodies) {
+		case 1:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-plan","function":{"name":"plan","arguments":"{\"steps\":[{\"task\":\"create tasks\"},{\"task\":\"notify owner\"}]}"}}]}}]}`))
+		case 2:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-add","function":{"name":"task_TaskService_Add","arguments":"{\"title\":\"Design\"}"}}]}}]}`))
+		case 3:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-delegate","function":{"name":"delegate","arguments":"{\"task\":\"notify owner@acme.com\",\"to\":\"comms\"}"}}]}}]}`))
+		case 4:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"done"}}]}`))
+		default:
+			t.Fatalf("unexpected API call %d", len(bodies))
+		}
+	}))
+	defer ts.Close()
+
+	var calls []string
+	p := NewProvider(
+		ai.WithAPIKey("test-key"),
+		ai.WithBaseURL(ts.URL),
+		ai.WithToolHandler(func(ctx context.Context, call ai.ToolCall) ai.ToolResult {
+			calls = append(calls, call.Name)
+			return ai.ToolResult{ID: call.ID, Content: `{"ok":true}`}
+		}),
+	)
+	resp, err := p.Generate(context.Background(), &ai.Request{
+		Prompt: "plan, create tasks, and delegate notification",
+		Tools: []ai.Tool{
+			{Name: "plan", Description: "record a plan", Properties: map[string]any{"steps": map[string]any{"type": "array"}}},
+			{Name: "task_TaskService_Add", Description: "add task", Properties: map[string]any{"title": map[string]any{"type": "string"}}},
+			{Name: "delegate", Description: "delegate work", Properties: map[string]any{"task": map[string]any{"type": "string"}, "to": map[string]any{"type": "string"}}},
+		},
+	})
+	if err != nil {
+		t.Fatalf("Generate returned error: %v", err)
+	}
+	wantCalls := []string{"plan", "task_TaskService_Add", "delegate"}
+	if strings.Join(calls, ",") != strings.Join(wantCalls, ",") {
+		t.Fatalf("tool calls = %v, want %v", calls, wantCalls)
+	}
+	if len(resp.ToolCalls) != 3 {
+		t.Fatalf("ToolCalls = %+v, want all multi-step calls", resp.ToolCalls)
+	}
+	if resp.Answer != "done" {
+		t.Fatalf("Answer = %q, want final follow-up reply", resp.Answer)
+	}
+	if len(bodies) != 4 {
+		t.Fatalf("requests = %d, want initial plus three follow-ups", len(bodies))
+	}
+	for i := 1; i < 4; i++ {
+		if _, ok := bodies[i]["tools"].([]any); !ok {
+			t.Fatalf("follow-up request %d did not include tools: %#v", i+1, bodies[i])
+		}
+	}
+}
+
+func TestProvider_GeneratePreservesFollowUpTextToolCallInReply(t *testing.T) {
+	var bodies []map[string]any
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		var body map[string]any
+		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		bodies = append(bodies, body)
+		w.Header().Set("Content-Type", "application/json")
+		switch len(bodies) {
+		case 1:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-1","function":{"name":"conformance_echo","arguments":"{\"value\":\"agent-conformance\"}"}}]}}]}`))
+		case 2:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"<tool_call name=\"delegate\">{\"task\":\"summarize the conformance marker\",\"to\":\"blocked-reviewer\"}</tool_call>"}}]}`))
+		default:
+			t.Fatalf("unexpected API call %d", len(bodies))
+		}
+	}))
+	defer ts.Close()
+
+	p := NewProvider(
+		ai.WithAPIKey("test-key"),
+		ai.WithBaseURL(ts.URL),
+		ai.WithToolHandler(func(ctx context.Context, call ai.ToolCall) ai.ToolResult {
+			if call.Name != "conformance_echo" {
+				t.Fatalf("unexpected structured tool call %+v", call)
+			}
+			return ai.ToolResult{ID: call.ID, Content: `{"marker":"agent-conformance-ok"}`}
+		}),
+	)
+	resp, err := p.Generate(context.Background(), &ai.Request{
+		Prompt: "run conformance",
+		Tools: []ai.Tool{
+			{Name: "conformance_echo", Description: "echo conformance marker", Properties: map[string]any{"value": map[string]any{"type": "string"}}},
+			{Name: "delegate", Description: "delegate work", Properties: map[string]any{"task": map[string]any{"type": "string"}, "to": map[string]any{"type": "string"}}},
+		},
+	})
+	if err != nil {
+		t.Fatalf("Generate returned error: %v", err)
+	}
+	if !strings.Contains(resp.Reply, `<tool_call name="delegate">`) {
+		t.Fatalf("Reply = %q, want tagged delegate follow-up for agent text fallback", resp.Reply)
+	}
+	if resp.Answer != "" {
+		t.Fatalf("Answer = %q, want follow-up text preserved only as Reply", resp.Answer)
+	}
+}
+
+func TestProvider_GenerateRetriesMinimaxBuiltInsAsTextTools(t *testing.T) {
+	var bodies []map[string]any
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		var body map[string]any
+		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		bodies = append(bodies, body)
+		w.Header().Set("Content-Type", "application/json")
+		switch len(bodies) {
+		case 1:
+			http.Error(w, `{"code":400,"msg":"bad request"}`, http.StatusBadRequest)
+		case 2:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"<tool_call name=\"delegate\">{\"task\":\"summarize\",\"to\":\"blocked-reviewer\"}</tool_call>"}}]}`))
+		default:
+			t.Fatalf("unexpected API call %d", len(bodies))
+		}
+	}))
+	defer ts.Close()
+
+	p := NewProvider(ai.WithAPIKey("test-key"), ai.WithBaseURL(ts.URL), ai.WithModel("minimaxai/minimax-m3"))
+	resp, err := p.Generate(context.Background(), &ai.Request{
+		Prompt: "plan and delegate",
+		Tools: []ai.Tool{
+			{Name: "task_TaskService_Add", Description: "add task", Properties: map[string]any{"title": map[string]any{"type": "string"}}},
+			{Name: "plan", Description: "record a plan", Properties: map[string]any{"steps": map[string]any{"type": "array"}}},
+			{Name: "request_input", Description: "request input", Properties: map[string]any{"prompt": map[string]any{"type": "string"}}},
+			{Name: "delegate", Description: "delegate work", Properties: map[string]any{"task": map[string]any{"type": "string"}, "to": map[string]any{"type": "string"}}},
+		},
+	})
+	if err != nil {
+		t.Fatalf("Generate returned error: %v", err)
+	}
+	if !strings.Contains(resp.Reply, `<tool_call name="delegate">`) {
+		t.Fatalf("Reply = %q, want text delegate fallback", resp.Reply)
+	}
+	if len(bodies) != 2 {
+		t.Fatalf("requests = %d, want initial plus compat retry", len(bodies))
+	}
+	initialTools := bodies[0]["tools"].([]any)
+	if len(initialTools) != 4 {
+		t.Fatalf("initial tools = %d, want all tools", len(initialTools))
+	}
+	retryTools := bodies[1]["tools"].([]any)
+	if len(retryTools) != 1 {
+		t.Fatalf("retry tools = %d, want only service tools", len(retryTools))
+	}
+	fn := retryTools[0].(map[string]any)["function"].(map[string]any)
+	if fn["name"] != "task_TaskService_Add" {
+		t.Fatalf("retry tool name = %v, want service tool only", fn["name"])
+	}
+	msgs := bodies[1]["messages"].([]any)
+	compat := msgs[len(msgs)-1].(map[string]any)
+	if compat["role"] != "system" || !strings.Contains(compat["content"].(string), `<tool_call name="tool_name">`) {
+		t.Fatalf("compat instruction = %#v", compat)
+	}
+}
+
+func TestProvider_GenerateFollowUpRetriesWithoutToolsOnBadRequest(t *testing.T) {
+	var bodies []map[string]any
+	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		var body map[string]any
+		if err := json.NewDecoder(r.Body).Decode(&body); err != nil {
+			t.Fatalf("decode request: %v", err)
+		}
+		bodies = append(bodies, body)
+		w.Header().Set("Content-Type", "application/json")
+		switch len(bodies) {
+		case 1:
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"","tool_calls":[{"id":"call-1","function":{"name":"conformance_echo","arguments":"{\"value\":\"agent-conformance\"}"}}]}}]}`))
+		case 2:
+			http.Error(w, `{"code":400,"msg":"bad request"}`, http.StatusBadRequest)
+		case 3:
+			if _, ok := body["tools"]; ok {
+				t.Fatalf("no-tools retry still included tools: %#v", body["tools"])
+			}
+			_, _ = w.Write([]byte(`{"choices":[{"message":{"content":"done"}}]}`))
+		default:
+			t.Fatalf("unexpected API call %d", len(bodies))
+		}
+	}))
+	defer ts.Close()
+
+	var toolCalls int
+	p := NewProvider(
+		ai.WithAPIKey("test-key"),
+		ai.WithBaseURL(ts.URL),
+		ai.WithModel("minimaxai/minimax-m3"),
+		ai.WithToolHandler(func(ctx context.Context, call ai.ToolCall) ai.ToolResult {
+			toolCalls++
+			return ai.ToolResult{ID: call.ID, Content: `{"marker":"agent-conformance-ok"}`}
+		}),
+	)
+	resp, err := p.Generate(context.Background(), &ai.Request{
+		Prompt: "call a tool",
+		Tools:  []ai.Tool{{Name: "conformance_echo", Description: "echo", Properties: map[string]any{"value": map[string]any{"type": "string"}}}},
+	})
+	if err != nil {
+		t.Fatalf("Generate returned error: %v", err)
+	}
+	if resp.Answer != "done" {
+		t.Fatalf("Answer = %q, want done", resp.Answer)
+	}
+	if toolCalls != 1 {
+		t.Fatalf("tool handler calls = %d, want one (no duplicate side effect)", toolCalls)
+	}
+	if len(bodies) != 3 {
+		t.Fatalf("requests = %d, want chat, failed follow-up, no-tools follow-up", len(bodies))
+	}
+	if _, ok := bodies[1]["tools"]; !ok {
+		t.Fatalf("first follow-up did not include tools")
+	}
+}
+
 func TestProvider_GenerateToolCallHTTPErrorIncludesRequestContext(t *testing.T) {
 	ts := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		http.Error(w, `{"code":400,"msg":"bad request"}`, http.StatusBadRequest)

cmd/micro/README.md

@@ -66,10 +66,11 @@ provider-free agent path:
 
 ```
 micro agent demo
+micro examples
 ```
 
-That points at the smallest mock-model first-agent example and the no-secret
-transcript before you add provider-backed chat.
+Those commands point at the smallest mock-model first-agent example, the no-secret
+transcript, and the support app before you add provider-backed chat.
 
 ### Output
 

cmd/micro/cli/cli.go

@@ -46,6 +46,34 @@ Full local contract:
 
 Guide: https://go-micro.dev/docs/guides/zero-to-hero.html`
 
+const examplesWayfinding = `First-agent examples (no provider key required)
+
+Run these from a go-micro repository checkout in this order. For the complete
+examples map, open examples/INDEX.md:
+
+  1. Smallest service-backed agent
+       go run ./examples/first-agent
+     Proves an agent can call a service tool with the deterministic mock model.
+
+  2. No-secret support-agent transcript
+       go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentTranscript -count=1
+     Exercises service tools, mock-model chat, and inspectable run history.
+
+  3. Full services → agents → workflows reference app
+       go run ./examples/support
+     Shows the support desk service, agent, workflow, and approval gate together.
+
+Then continue the same path with the installed CLI:
+  micro agent demo
+  micro docs
+  micro zero-to-hero
+
+Guides:
+  https://go-micro.dev/docs/guides/no-secret-first-agent.html
+  https://go-micro.dev/docs/guides/your-first-agent.html
+  https://go-micro.dev/docs/guides/debugging-agents.html
+  https://go-micro.dev/docs/guides/zero-to-hero.html`
+
 const docsWayfinding = `First-agent and 0→hero docs:
 
   1. Start with the no-secret CLI demo
@@ -150,6 +178,17 @@ func init() {
 				return nil
 			},
 		},
+
+		{
+			Name:  "examples",
+			Usage: "Show provider-free first-agent example paths",
+			Description: `Print the maintained no-secret examples for the services → agents →
+workflows on-ramp: first-agent, transcript, support app, and matching guides.`,
+			Action: func(ctx *cli.Context) error {
+				fmt.Fprintln(ctx.App.Writer, examplesWayfinding)
+				return nil
+			},
+		},
 		{
 			Name:  "zero-to-hero",
 			Usage: "Show the no-secret 0→hero lifecycle demo command",

cmd/micro/examples_wayfinding_test.go

@@ -0,0 +1,71 @@
+package main
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/urfave/cli/v2"
+	microcmd "go-micro.dev/v6/cmd"
+)
+
+func TestExamplesWayfindingIndexStaysLinked(t *testing.T) {
+	root := filepath.Join("..", "..")
+	files := map[string]string{}
+	for _, name := range []string{"README.md", "examples/README.md", "examples/INDEX.md"} {
+		b, err := os.ReadFile(filepath.Join(root, filepath.FromSlash(name)))
+		if err != nil {
+			t.Fatalf("read %s: %v", name, err)
+		}
+		files[name] = string(b)
+	}
+
+	for _, check := range []struct {
+		file string
+		want []string
+	}{
+		{
+			file: "README.md",
+			want: []string{"examples/INDEX.md", "examples/first-agent/", "examples/support/", "zero-to-hero.md"},
+		},
+		{
+			file: "examples/README.md",
+			want: []string{"./INDEX.md", "./first-agent/", "./support/", "./mcp/hello/", "./mcp/workflow/"},
+		},
+		{
+			file: "examples/INDEX.md",
+			want: []string{"go run ./examples/first-agent", "go run ./examples/support", "mcp/hello", "mcp/workflow", "flow-durable", "micro examples"},
+		},
+	} {
+		for _, want := range check.want {
+			if !strings.Contains(files[check.file], want) {
+				t.Fatalf("%s missing %q", check.file, want)
+			}
+		}
+	}
+}
+
+func TestExamplesCommandPointsAtWayfindingIndex(t *testing.T) {
+	examples := commandByName(t, "examples")
+	var out bytes.Buffer
+	app := cli.NewApp()
+	app.Writer = &out
+	if err := examples.Action(cli.NewContext(app, nil, nil)); err != nil {
+		t.Fatalf("micro examples failed: %v", err)
+	}
+
+	for _, want := range []string{
+		"examples/INDEX.md",
+		"go run ./examples/first-agent",
+		"go run ./examples/support",
+		"micro zero-to-hero",
+	} {
+		if !strings.Contains(out.String(), want) {
+			t.Fatalf("micro examples output missing %q:\n%s", want, out.String())
+		}
+	}
+
+	_ = microcmd.DefaultCmd // keep this test coupled to the registered command package.
+}

cmd/micro/first_agent_walkthrough_test.go

@@ -22,7 +22,7 @@ func TestFirstAgentWalkthroughCLIBoundaries(t *testing.T) {
 		}
 	}
 
-	for _, want := range []string{"new", "run", "chat", "inspect", "agent", "docs"} {
+	for _, want := range []string{"new", "run", "chat", "inspect", "agent", "docs", "examples"} {
 		if !commands[want] {
 			t.Fatalf("first-agent walkthrough missing %q command", want)
 		}
@@ -79,6 +79,32 @@ func TestFirstAgentWalkthroughCLIBoundaries(t *testing.T) {
 		t.Fatalf("micro docs output should use the first-agent inspect command, not the legacy runs shortcut:\n%s", out.String())
 	}
 
+	examples := commandByName(t, "examples")
+	if !strings.Contains(examples.Usage, "first-agent") {
+		t.Fatalf("micro examples should advertise the first-agent examples path; usage was %q", examples.Usage)
+	}
+	out.Reset()
+	if err := examples.Action(cli.NewContext(app, nil, nil)); err != nil {
+		t.Fatalf("micro examples failed: %v", err)
+	}
+	for _, want := range []string{
+		"First-agent examples",
+		"go run ./examples/first-agent",
+		"go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentTranscript -count=1",
+		"go run ./examples/support",
+		"micro agent demo",
+		"micro docs",
+		"micro zero-to-hero",
+		"no-secret-first-agent.html",
+		"your-first-agent.html",
+		"debugging-agents.html",
+		"zero-to-hero.html",
+	} {
+		if !strings.Contains(out.String(), want) {
+			t.Fatalf("micro examples output missing %q:\n%s", want, out.String())
+		}
+	}
+
 	agent := commandByName(t, "agent")
 	if !strings.Contains(agent.Usage, "micro agent demo") {
 		t.Fatalf("micro agent help should advertise the no-secret demo; usage was %q", agent.Usage)

examples/INDEX.md

@@ -0,0 +1,46 @@
+# Examples wayfinding
+
+Use this index when you want the shortest path from a first runnable agent to the
+next services, agents, workflows, and interop examples. Every command below is
+provider-free unless the example README says otherwise.
+
+## Pick by goal
+
+| Goal | Start here | Run or verify | Then try |
+|------|------------|---------------|----------|
+| Run the smallest no-secret agent | [`first-agent`](./first-agent/) | `go run ./examples/first-agent` | [`agent-demo`](./agent-demo/) for a larger service-backed agent |
+| Prove the maintained 0→hero path | [`support`](./support/) | `go run ./examples/support` and `go test ./examples/support` | [`zero-to-hero` guide](../internal/website/docs/guides/zero-to-hero.md) |
+| See planning and delegation | [`agent-plan-delegate`](./agent-plan-delegate/) | `go run ./examples/agent-plan-delegate` | [`plan-delegate` guide](../internal/website/docs/guides/plan-delegate.md) |
+| Expose services through MCP | [`mcp/hello`](./mcp/hello/) | follow [`mcp`](./mcp/) setup | [`mcp/crud`](./mcp/crud/) and [`mcp/workflow`](./mcp/workflow/) |
+| Try A2A or gRPC interop next | [`agent-demo`](./agent-demo/) plus gateway docs | run the example, then use the gateway docs | [`grpc-interop`](./grpc-interop/) |
+| Add workflow durability | [`flow-durable`](./flow-durable/) | `go run ./examples/flow-durable` | [`flow-loop`](./flow-loop/) |
+
+## Recommended adoption path
+
+1. **First service:** run [`hello-world`](./hello-world/) to learn service
+   registration, handlers, client calls, and health checks.
+2. **First agent:** run [`first-agent`](./first-agent/) with
+   `go run ./examples/first-agent`; it uses a deterministic mock model and needs
+   no provider key.
+3. **0→hero reference:** run [`support`](./support/) with
+   `go run ./examples/support`; it keeps typed services, an agent chat loop, an
+   event-driven flow, and an approval gate in one maintained example.
+4. **Interop next:** use [`mcp/hello`](./mcp/hello/), [`mcp/crud`](./mcp/crud/),
+   and [`mcp/workflow`](./mcp/workflow/) when you are ready to expose tools to
+   external AI clients.
+5. **Workflow depth:** use [`flow-durable`](./flow-durable/) once the agent path
+   needs checkpointed, resumable deterministic work.
+
+## CLI wayfinding
+
+The installed CLI prints the same path:
+
+```bash
+micro examples
+micro agent demo
+micro zero-to-hero
+```
+
+Keep this file, [`README.md`](../README.md), and the `micro examples` output in
+sync so new developers can find `examples/first-agent` and `examples/support`
+from one documented path.

examples/README.md

@@ -7,8 +7,8 @@ coordinate work with workflows.
 ## Quick Start
 
 Each example can be run with `go run .` from its directory unless its README says
-otherwise. If you are new to the repo, follow the first-agent path below instead
-of reading the directories alphabetically.
+otherwise. If you are new to the repo, start with the [examples wayfinding index](./INDEX.md)
+or follow the first-agent path below instead of reading the directories alphabetically.
 
 ## Recommended first-agent path
 

internal/harness/install-smoke/run.sh

@@ -49,6 +49,34 @@ require_output() {
   fi
 }
 
+require_ordered_output() {
+  local description=$1
+  shift
+  local -a expected=()
+  while [[ $# -gt 0 && "$1" != "--" ]]; do
+    expected+=("$1")
+    shift
+  done
+  shift
+
+  local output
+  if ! output=$("$MICRO" "$@" 2>&1); then
+    echo "micro $* failed while checking $description" >&2
+    echo "$output" >&2
+    exit 1
+  fi
+
+  local remainder=$output
+  for text in "${expected[@]}"; do
+    if [[ "$remainder" != *"$text"* ]]; then
+      echo "micro $* missing expected ordered text '$text' for $description" >&2
+      echo "$output" >&2
+      exit 1
+    fi
+    remainder=${remainder#*"$text"}
+  done
+}
+
 require_output "version" "micro version" --version
 require_output "root help" "COMMANDS" --help
 require_output "service scaffold" "micro new" new --help
@@ -58,4 +86,51 @@ require_output "agent chat" "micro chat" chat --help
 require_output "agent inspection" "micro inspect agent" inspect agent --help
 require_output "flow inspection" "micro inspect flow" inspect flow --help
 
+require_ordered_output "installed first-agent docs wayfinding" \
+  "micro agent demo" \
+  "no-secret-first-agent.html" \
+  "your-first-agent.html" \
+  "micro agent preflight  # before micro run: prerequisites" \
+  "micro run" \
+  "micro chat" \
+  "micro agent doctor     # after micro run: chat/gateway/inspect recovery" \
+  "debugging-agents.html" \
+  "micro inspect agent <name>" \
+  "zero-to-hero.html" \
+  -- docs
+
+require_ordered_output "installed provider-free examples wayfinding" \
+  "go run ./examples/first-agent" \
+  "go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentTranscript -count=1" \
+  "go run ./examples/support" \
+  "micro agent demo" \
+  "micro docs" \
+  "micro zero-to-hero" \
+  "no-secret-first-agent.html" \
+  "your-first-agent.html" \
+  "debugging-agents.html" \
+  "zero-to-hero.html" \
+  -- examples
+
+require_ordered_output "installed no-secret agent demo" \
+  "provider-free" \
+  "go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentTranscript -count=1" \
+  "your-first-agent.html" \
+  "debugging-agents.html" \
+  "zero-to-hero.html" \
+  "micro agent preflight  # before micro run: prerequisites" \
+  "micro run" \
+  "micro chat" \
+  "micro agent doctor     # after micro run: chat/gateway/inspect recovery" \
+  "micro inspect agent <name>" \
+  -- agent demo
+
+require_ordered_output "installed zero-to-hero lifecycle wayfinding" \
+  "./internal/harness/zero-to-hero-ci/run.sh" \
+  "go run ./examples/first-agent" \
+  "go run ./examples/support" \
+  "make harness" \
+  "zero-to-hero.html" \
+  -- zero-to-hero
+
 echo "✓ install smoke path verified"

internal/harness/plan-delegate/main.go

@@ -157,7 +157,7 @@ func (s *NotifyService) Send(ctx context.Context, req *SendRequest, rsp *SendRes
 	if s.bySend == nil {
 		s.bySend = map[string]bool{}
 	}
-	key := strings.ToLower(strings.TrimSpace(req.To)) + "\x00" + strings.ToLower(strings.TrimSpace(req.Message))
+	key := notifyDedupKey(req.To, req.Message)
 	s.attempts++
 	if !s.bySend[key] {
 		s.bySend[key] = true
@@ -184,6 +184,51 @@ func (s *NotifyService) duplicateAttempts() int {
 	return s.duplicates
 }
 
+func notifyDedupKey(to, message string) string {
+	recipient := canonicalLaunchNotifyRecipient(normalizeNotifyText(to))
+	body := normalizeNotifyText(message)
+	if isLaunchReadinessNotify(body) {
+		body = "launch-readiness"
+	}
+	return recipient + "\x00" + body
+}
+
+func canonicalLaunchNotifyRecipient(recipient string) string {
+	switch recipient {
+	case "owner", "launch owner", "plan owner", "owner acme com", "owner@acme com", "owner @ acme com":
+		return "owner@acme.com"
+	default:
+		if strings.Contains(recipient, "owner") && strings.Contains(recipient, "acme") {
+			return "owner@acme.com"
+		}
+		return recipient
+	}
+}
+
+func normalizeNotifyText(message string) string {
+	message = strings.ToLower(strings.TrimSpace(message))
+	message = strings.Map(func(r rune) rune {
+		switch {
+		case r >= 'a' && r <= 'z', r >= '0' && r <= '9':
+			return r
+		case r == '@':
+			return r
+		default:
+			return ' '
+		}
+	}, message)
+	return strings.Join(strings.Fields(message), " ")
+}
+
+func isLaunchReadinessNotify(message string) bool {
+	return strings.Contains(message, "launch") &&
+		strings.Contains(message, "plan") &&
+		(strings.Contains(message, "ready") ||
+			strings.Contains(message, "readiness") ||
+			strings.Contains(message, "prepared") ||
+			strings.Contains(message, "complete"))
+}
+
 // ---------------------------------------------------------------------------
 // mock LLM provider — the ONLY fake. It "reasons" by simple heuristics
 // over the tools it's offered and the system prompt it's given, calling
@@ -203,6 +248,11 @@ type mockModel struct {
 	// duplicateNotify makes the comms mock replay the same notification call.
 	// The notify service should collapse that replay to one durable side effect.
 	duplicateNotify bool
+
+	// duplicateDelegate makes the conductor mock replay the same delegate call.
+	// The delegate idempotency path should collapse that replay before it can
+	// ask the delegated comms agent to notify twice.
+	duplicateDelegate bool
 }
 
 func newMock(opts ...ai.Option) ai.Model {
@@ -223,6 +273,12 @@ func newMockDuplicateNotify(opts ...ai.Option) ai.Model {
 	return m
 }
 
+func newMockDuplicateDelegate(opts ...ai.Option) ai.Model {
+	m := &mockModel{duplicateDelegate: true}
+	_ = m.Init(opts...)
+	return m
+}
+
 func (m *mockModel) Init(opts ...ai.Option) error {
 	for _, o := range opts {
 		o(&m.opts)
@@ -299,10 +355,14 @@ func (m *mockModel) Generate(ctx context.Context, req *ai.Request, _ ...ai.Gener
 					"to":   "comms",
 				})
 			} else {
-				m.call("conductor", del, map[string]any{
+				input := map[string]any{
 					"task": delegatedNotifyTask,
 					"to":   "comms",
-				})
+				}
+				m.call("conductor", del, input)
+				if m.duplicateDelegate {
+					m.call("conductor", del, input)
+				}
 			}
 		}
 		return &ai.Response{Answer: "Created Design, Build and Ship, and had comms notify the owner."}, nil
@@ -338,6 +398,8 @@ func runPlanDelegate(provider string) error {
 		ai.Register("mock-unknown-delegate", newMockUnknownDelegate)
 	case "mock-duplicate-notify":
 		ai.Register("mock-duplicate-notify", newMockDuplicateNotify)
+	case "mock-duplicate-delegate":
+		ai.Register("mock-duplicate-delegate", newMockDuplicateDelegate)
 	default:
 		apiKey = providerKey(provider)
 		if apiKey == "" {
@@ -553,7 +615,7 @@ func isClientTimeout(err error) bool {
 }
 
 func main() {
-	provider := flag.String("provider", "mock", "LLM provider: mock (default), mock-unknown-delegate, mock-duplicate-notify, anthropic, openai, gemini, groq, mistral, together, atlascloud")
+	provider := flag.String("provider", "mock", "LLM provider: mock (default), mock-unknown-delegate, mock-duplicate-notify, mock-duplicate-delegate, anthropic, openai, gemini, groq, mistral, together, atlascloud")
 	flag.Parse()
 
 	if err := runPlanDelegate(*provider); err != nil {

internal/harness/plan-delegate/main_test.go

@@ -240,6 +240,15 @@ func TestPlanDelegateIdempotentDuplicateNotifyReplay(t *testing.T) {
 	}
 }
 
+func TestPlanDelegateIdempotentDuplicateDelegateReplay(t *testing.T) {
+	if testing.Short() {
+		t.Skip("0→hero harness boots an end-to-end system; skipped with -short")
+	}
+	if err := runPlanDelegate("mock-duplicate-delegate"); err != nil {
+		t.Fatalf("0→hero harness with duplicate delegate replay: %v", err)
+	}
+}
+
 func TestTaskServiceAddIsIdempotentForLaunchTitles(t *testing.T) {
 	svc := new(TaskService)
 	for _, title := range []string{"Design", "design task", "Build", "Build launch task", "Ship", "ship readiness"} {
@@ -422,9 +431,18 @@ func TestPlanDelegateExecutionClassifiesPartialClientTimeout(t *testing.T) {
 
 func TestNotifyServiceSendIsIdempotentForDuplicateDelivery(t *testing.T) {
 	svc := new(NotifyService)
-	for i := 0; i < 3; i++ {
+	messages := []string{
+		"The launch plan is ready",
+		"The launch plan is ready.",
+		"Launch readiness: the plan is ready!",
+	}
+	for i, message := range messages {
 		var rsp SendResponse
-		if err := svc.Send(context.Background(), &SendRequest{To: "owner@acme.com", Message: "The launch plan is ready"}, &rsp); err != nil {
+		to := "owner@acme.com"
+		if i == len(messages)-1 {
+			to = "owner"
+		}
+		if err := svc.Send(context.Background(), &SendRequest{To: to, Message: message}, &rsp); err != nil {
 			t.Fatalf("Send attempt %d: %v", i+1, err)
 		}
 		if !rsp.Sent {
@@ -434,4 +452,32 @@ func TestNotifyServiceSendIsIdempotentForDuplicateDelivery(t *testing.T) {
 	if got := svc.count(); got != 1 {
 		t.Fatalf("notify count = %d, want 1 after duplicate delivery replays", got)
 	}
+	if got := svc.duplicateAttempts(); got != len(messages)-1 {
+		t.Fatalf("duplicate notify attempts = %d, want %d", got, len(messages)-1)
+	}
+}
+
+func TestNotifyServiceCollapsesProviderReadinessParaphrases(t *testing.T) {
+	svc := new(NotifyService)
+	requests := []SendRequest{
+		{To: "owner@acme.com", Message: "The launch plan is ready"},
+		{To: "owner @ acme.com", Message: "Launch plan ready."},
+		{To: "launch owner", Message: "The launch readiness plan is prepared."},
+		{To: "plan owner", Message: "Launch plan is complete!"},
+	}
+	for i, req := range requests {
+		var rsp SendResponse
+		if err := svc.Send(context.Background(), &req, &rsp); err != nil {
+			t.Fatalf("Send attempt %d: %v", i+1, err)
+		}
+		if !rsp.Sent {
+			t.Fatalf("Send attempt %d reported Sent=false", i+1)
+		}
+	}
+	if got := svc.count(); got != 1 {
+		t.Fatalf("notify count = %d, want 1 after provider paraphrase replays", got)
+	}
+	if got := svc.duplicateAttempts(); got != len(requests)-1 {
+		t.Fatalf("duplicate notify attempts = %d, want %d", got, len(requests)-1)
+	}
 }

internal/harness/zero-to-hero-ci/README.md

@@ -11,9 +11,11 @@ scripted so CI can run it on every push without external services or model keys.
    documented first-agent walkthrough path.
 2. **Run** — `micro run` remains available as the local development entry point.
 3. **Chat** — `micro chat` remains available as the interactive agent entry point.
-4. **Inspect** — `micro inspect agent <name>` and `micro inspect flow <name>`
-   remain available as the local run-history inspection step, with `micro flow
-   runs` preserving durable workflow history inspection.
+4. **Inspect/debugging** — `micro inspect agent <name>`, `micro agent history <name>`,
+   and `micro inspect flow <name>` remain available as the local run-history
+   inspection step. The no-secret debugging smoke seeds durable agent run history
+   and memory, then runs the documented inspect/history commands without provider
+   credentials; `micro flow runs` preserves durable workflow history inspection.
 5. **Deploy** — `micro deploy --dry-run <target>` remains available as the
    deployment-boundary checkpoint. The dry run resolves configured deploy targets
    and services and prints the remote build/copy/systemd/health plan without

internal/harness/zero-to-hero-ci/docs_test.go

@@ -1,11 +1,17 @@
 package zerotoheroci
 
 import (
+	"encoding/json"
 	"os"
+	"os/exec"
 	"path/filepath"
 	"regexp"
 	"strings"
 	"testing"
+	"time"
+
+	goagent "go-micro.dev/v6/agent"
+	"go-micro.dev/v6/store"
 )
 
 func TestZeroToHeroReferenceDocs(t *testing.T) {
@@ -77,6 +83,76 @@ func TestGuidesNavigationLeadsWithDoing(t *testing.T) {
 	}
 }
 
+func TestYourFirstAgentTutorialSmoke(t *testing.T) {
+	root := filepath.Clean(filepath.Join("..", "..", ".."))
+	absRoot, err := filepath.Abs(root)
+	if err != nil {
+		t.Fatalf("resolve repository root: %v", err)
+	}
+	guide := readFile(t, filepath.Join(root, "internal", "website", "docs", "guides", "your-first-agent.md"))
+
+	for _, want := range []string{
+		"go test ./internal/harness/zero-to-hero-ci -run TestYourFirstAgentTutorialSmoke -count=1",
+		"micro agent preflight",
+		"mkdir first-agent",
+		"go mod init example.com/first-agent",
+		"go get go-micro.dev/v6@v6",
+		"micro run",
+		"micro call task TaskService.Create",
+		"micro call task TaskService.List",
+		"micro chat assistant",
+		"micro inspect agent assistant",
+	} {
+		if !strings.Contains(guide, want) {
+			t.Fatalf("Your First Agent guide missing copy/paste boundary %q", want)
+		}
+	}
+
+	mainGo := extractFirstAgentMain(t, guide)
+	workspace := t.TempDir()
+	writeFile(t, filepath.Join(workspace, "go.mod"), "module example.com/first-agent\n\ngo 1.24\n\nrequire go-micro.dev/v6 v6.0.0\n\nreplace go-micro.dev/v6 => "+absRoot+"\n")
+	writeFile(t, filepath.Join(workspace, "main.go"), mainGo)
+
+	runInWorkspace(t, workspace, "go", "mod", "tidy")
+	runInWorkspace(t, workspace, "go", "test", "./...")
+}
+
+func extractFirstAgentMain(t *testing.T, guide string) string {
+	t.Helper()
+	start := strings.Index(guide, "Add `main.go`:")
+	if start == -1 {
+		t.Fatal("Your First Agent guide is missing the main.go section")
+	}
+	rest := guide[start:]
+	open := strings.Index(rest, "```go")
+	if open == -1 {
+		t.Fatal("Your First Agent guide is missing a Go code fence for main.go")
+	}
+	rest = rest[open+len("```go"):]
+	close := strings.Index(rest, "```")
+	if close == -1 {
+		t.Fatal("Your First Agent guide main.go code fence is not closed")
+	}
+	return strings.TrimSpace(rest[:close]) + "\n"
+}
+
+func writeFile(t *testing.T, name, contents string) {
+	t.Helper()
+	if err := os.WriteFile(name, []byte(contents), 0o644); err != nil {
+		t.Fatalf("write %s: %v", name, err)
+	}
+}
+
+func runInWorkspace(t *testing.T, workspace, name string, args ...string) {
+	t.Helper()
+	cmd := exec.Command(name, args...)
+	cmd.Dir = workspace
+	out, err := cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("Your First Agent tutorial command %q does not pass from a clean workspace: %v\n%s", strings.Join(append([]string{name}, args...), " "), err, out)
+	}
+}
+
 func TestArchitectureDocsAlignWithAgentHarnessLifecycle(t *testing.T) {
 	root := filepath.Clean(filepath.Join("..", "..", ".."))
 	doc := readFile(t, filepath.Join(root, "internal", "website", "docs", "architecture.md"))
@@ -172,6 +248,40 @@ func TestFirstAgentWayfindingDocs(t *testing.T) {
 				"guides/install-troubleshooting.html",
 				"micro agent demo",
 				"micro zero-to-hero",
+				"https://github.com/micro/go-micro/blob/master/examples/INDEX.md",
+				"https://github.com/micro/go-micro/tree/master/examples/support",
+				"https://github.com/micro/go-micro/tree/master/examples/first-agent",
+				"guides/no-secret-first-agent.html",
+				"guides/your-first-agent.html",
+				"guides/debugging-agents.html",
+				"guides/zero-to-hero.html",
+			},
+		},
+		{
+			name:    "website quickstart next steps",
+			file:    filepath.Join(root, "internal", "website", "docs", "quickstart.md"),
+			heading: "## Next Steps",
+			links: []string{
+				"guides/install-troubleshooting.html",
+				"micro agent demo",
+				"micro zero-to-hero",
+				"https://github.com/micro/go-micro/blob/master/examples/INDEX.md",
+				"https://github.com/micro/go-micro/tree/master/examples/support",
+				"https://github.com/micro/go-micro/tree/master/examples/first-agent",
+				"guides/no-secret-first-agent.html",
+				"guides/your-first-agent.html",
+				"guides/debugging-agents.html",
+				"guides/zero-to-hero.html",
+			},
+		},
+		{
+			name:    "website docs index learn more",
+			file:    filepath.Join(root, "internal", "website", "docs", "index.md"),
+			heading: "## Learn More",
+			links: []string{
+				"getting-started.html",
+				"https://github.com/micro/go-micro/blob/master/examples/INDEX.md",
+				"https://github.com/micro/go-micro/tree/master/examples/support",
 				"guides/no-secret-first-agent.html",
 				"guides/your-first-agent.html",
 				"guides/debugging-agents.html",
@@ -231,6 +341,16 @@ func TestFirstAgentWayfindingLinkTargetsResolve(t *testing.T) {
 			file:    filepath.Join(root, "internal", "website", "docs", "getting-started.md"),
 			heading: "### First-agent on-ramp",
 		},
+		{
+			name:    "website quickstart next steps",
+			file:    filepath.Join(root, "internal", "website", "docs", "quickstart.md"),
+			heading: "## Next Steps",
+		},
+		{
+			name:    "website docs index learn more",
+			file:    filepath.Join(root, "internal", "website", "docs", "index.md"),
+			heading: "## Learn More",
+		},
 	}
 
 	for _, check := range checks {
@@ -498,6 +618,110 @@ func TestNoSecretFirstAgentTranscript(t *testing.T) {
 	}
 }
 
+func TestNoSecretFirstAgentDebuggingSmoke(t *testing.T) {
+	root := filepath.Clean(filepath.Join("..", "..", ".."))
+	home := t.TempDir()
+	storeDir := filepath.Join(home, "micro", "store")
+	st := store.NewFileStore(store.DirOption(storeDir))
+
+	seedNoSecretAgentDebuggingState(t, st)
+	if err := st.Close(); err != nil {
+		t.Fatalf("close seeded store: %v", err)
+	}
+
+	micro := buildMicroBinary(t, root)
+
+	for _, tc := range []struct {
+		name string
+		args []string
+		want []string
+	}{
+		{
+			name: "demo advertises provider-free debug path",
+			args: []string{"agent", "demo"},
+			want: []string{"No-secret first-agent demo", "provider-free", "run history", "micro inspect agent <name>"},
+		},
+		{
+			name: "inspect shows seeded run history",
+			args: []string{"inspect", "agent", "assistant", "--limit", "1"},
+			want: []string{`Agent "assistant" runs`, "run-debug-smoke", "status=done", "events=3", "last=done", "trace=trace-debug-"},
+		},
+		{
+			name: "inspect filters documented statuses",
+			args: []string{"inspect", "agent", "--status", "done", "--json", "assistant"},
+			want: []string{"run-debug-smoke", `"status": "done"`, `"trace_id": "trace-debug-smoke"`},
+		},
+		{
+			name: "agent history shows memory and run index",
+			args: []string{"agent", "history", "assistant"},
+			want: []string{"user:", "Triage ticket-1", "assistant:", "ticket-1 is ready", "Runs:", "run-debug-smoke", "status=done"},
+		},
+	} {
+		t.Run(tc.name, func(t *testing.T) {
+			out := runMicroCLIWithHome(t, micro, home, tc.args...)
+			for _, want := range tc.want {
+				if !strings.Contains(out, want) {
+					t.Fatalf("micro %s output missing %q:\n%s", strings.Join(tc.args, " "), want, out)
+				}
+			}
+		})
+	}
+}
+
+func seedNoSecretAgentDebuggingState(t *testing.T, st store.Store) {
+	t.Helper()
+	scoped := store.Scope(st, "agent", "assistant")
+	runID := "run-debug-smoke"
+	events := []goagent.RunEvent{
+		{Time: time.Unix(1700000000, 0), RunID: runID, Agent: "assistant", TraceID: "trace-debug-smoke", Kind: "run", Name: "ask"},
+		{Time: time.Unix(1700000001, 0), RunID: runID, Agent: "assistant", TraceID: "trace-debug-smoke", Kind: "model", Provider: "mock", Model: "first-agent-mock"},
+		{Time: time.Unix(1700000002, 0), RunID: runID, Agent: "assistant", TraceID: "trace-debug-smoke", Kind: "done", Name: "answer"},
+	}
+	for _, event := range events {
+		b, err := json.Marshal(event)
+		if err != nil {
+			t.Fatal(err)
+		}
+		key := "runs/" + event.RunID + "/" + event.Time.Format("20060102150405.000000000") + "-" + event.Kind
+		if err := scoped.Write(&store.Record{Key: key, Value: b}); err != nil {
+			t.Fatalf("seed run event: %v", err)
+		}
+	}
+
+	mem := goagent.NewMemory(scoped, "history", 10)
+	mem.Add("user", "Triage ticket-1 for Alice")
+	mem.Add("assistant", "ticket-1 is ready for Alice without provider secrets")
+}
+
+func buildMicroBinary(t *testing.T, root string) string {
+	t.Helper()
+	bin := filepath.Join(t.TempDir(), "micro")
+	cmd := exec.Command("go", "build", "-o", bin, "./cmd/micro")
+	cmd.Dir = root
+	out, err := cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("build micro CLI failed: %v\n%s", err, out)
+	}
+	return bin
+}
+
+func runMicroCLIWithHome(t *testing.T, micro, home string, args ...string) string {
+	t.Helper()
+	cmd := exec.Command(micro, args...)
+	cmd.Env = append(os.Environ(),
+		"HOME="+home,
+		"MICRO_AI_API_KEY=",
+		"OPENAI_API_KEY=",
+		"ANTHROPIC_API_KEY=",
+		"GEMINI_API_KEY=",
+	)
+	out, err := cmd.CombinedOutput()
+	if err != nil {
+		t.Fatalf("micro %s failed: %v\n%s", strings.Join(args, " "), err, out)
+	}
+	return string(out)
+}
+
 func TestFirstAgentWayfindingTargetsExist(t *testing.T) {
 	root := filepath.Clean(filepath.Join("..", "..", ".."))
 	for _, target := range []string{

internal/harness/zero-to-hero-ci/run.sh

@@ -8,7 +8,7 @@ cd "$ROOT"
 # without secrets or long-running daemons.
 go test ./cmd/micro -run 'TestFirstAgentWalkthroughCLIBoundaries|TestZeroToHeroCLIBoundaries' -count=1
 go test ./cmd/micro/cli/deploy -run TestDeployDryRun -count=1
-go test ./internal/harness/zero-to-hero-ci -run 'TestNoSecretFirstAgentTranscript|TestZeroToHeroReferenceDocs' -count=1
+go test ./internal/harness/zero-to-hero-ci -run 'TestNoSecretFirstAgentTranscript|TestNoSecretFirstAgentDebuggingSmoke|TestZeroToHeroReferenceDocs|TestYourFirstAgentTutorialSmoke' -count=1
 
 # Deterministic no-secret reference scenarios. These use the real Go Micro
 # runtime and mock only the LLM provider. The support example is the maintained

internal/website/_data/navigation.yml

@@ -42,6 +42,8 @@ examples:
 guides:
   - title: Debugging your agent
     url: /docs/guides/debugging-agents.html
+  - title: micro loop quickstart
+    url: /docs/guides/micro-loop.html
   - title: Plan & Delegate
     url: /docs/guides/plan-delegate.html
   - title: Agent Guardrails
@@ -86,6 +88,7 @@ search_order:
   - /docs/guides/your-first-agent.html
   - /docs/guides/zero-to-hero.html
   - /docs/guides/debugging-agents.html
+  - /docs/guides/micro-loop.html
   - /docs/getting-started.html
   - /docs/mcp.html
   - /docs/architecture.html

internal/website/docs/getting-started.md

@@ -57,12 +57,14 @@ After this quick start, follow the agent path in order:
 
 1. [Install troubleshooting](guides/install-troubleshooting.html) — verify the CLI install before agent work.
 2. `micro agent demo` — print the provider-free first-agent demo command and next docs steps from the installed CLI.
-3. `micro zero-to-hero` — print the maintained one-command no-secret lifecycle harness and runnable examples.
-4. [Smallest first-agent example](https://github.com/micro/go-micro/tree/master/examples/first-agent) — run one service-backed agent with a mock model and no provider key.
-5. [No-secret first-agent transcript](guides/no-secret-first-agent.html) — run a useful support agent with a mock model before setting up a provider key.
-6. [Your First Agent](guides/your-first-agent.html) — build a service-backed agent and talk to it with `micro chat`.
-7. [Debugging your agent](guides/debugging-agents.html) — inspect service registration, tool calls, run history, memory, provider failures, and flow handoffs when the agent surprises you.
-8. [0→hero reference path](guides/zero-to-hero.html) — prove the full scaffold → run → chat → inspect → deploy dry-run lifecycle with commands exercised by `make harness`.
+3. `micro examples` — print the maintained provider-free runnable examples in copy/paste order.
+4. `micro zero-to-hero` — print the maintained one-command no-secret lifecycle harness and runnable examples.
+5. [Examples wayfinding index](https://github.com/micro/go-micro/blob/master/examples/INDEX.md) — choose the smallest no-secret first-agent, maintained [0→hero support reference](https://github.com/micro/go-micro/tree/master/examples/support), and next interop examples from one map.
+6. [Smallest first-agent example](https://github.com/micro/go-micro/tree/master/examples/first-agent) — run one service-backed agent with a mock model and no provider key.
+7. [No-secret first-agent transcript](guides/no-secret-first-agent.html) — run a useful support agent with a mock model before setting up a provider key.
+8. [Your First Agent](guides/your-first-agent.html) — build a service-backed agent and talk to it with `micro chat`.
+9. [Debugging your agent](guides/debugging-agents.html) — inspect service registration, tool calls, run history, memory, provider failures, and flow handoffs when the agent surprises you.
+10. [0→hero reference path](guides/zero-to-hero.html) — prove the full scaffold → run → chat → inspect → deploy dry-run lifecycle with commands exercised by `make harness`.
 
 ## Write a Service
 
@@ -258,4 +260,5 @@ The flow discovers all services as tools and lets the LLM decide which RPCs to c
 - [Agent Design](https://github.com/micro/go-micro/blob/master/internal/docs/AGENT_DESIGN.md) — the full agent interface specification
 - [MCP & AI Agents](mcp.html) — MCP gateway, tool discovery, and auth
 - [Data Model](model.html) — typed persistence with CRUD and queries
+- [`micro loop` quickstart](guides/micro-loop.html) — scaffold a CI-gated autonomous improvement loop for a repository
 - [Deployment](deployment.html) — deploy via SSH + systemd

internal/website/docs/guides/micro-loop.md

@@ -0,0 +1,96 @@
+---
+layout: default
+---
+
+# `micro loop` quickstart
+
+`micro loop` scaffolds the autonomous improvement loop that Go Micro uses on
+this repository: GitHub Actions workflows for planning, building, evaluation
+feedback, coherence, security, and release. Use it when you want a repository to
+continuously turn a ranked queue into small PRs while CI remains the merge gate.
+
+## 1. Initialize the loop
+
+Run the default loop from the repository root:
+
+```bash
+micro loop init
+```
+
+For every role used by Go Micro itself, scaffold all workflows:
+
+```bash
+micro loop init --roles all
+```
+
+The command writes:
+
+- `.github/loop/NORTH_STAR.md` — the direction every increment should optimize.
+- `.github/loop/PRIORITIES.md` — the ranked queue; the builder takes the top open issue.
+- `.github/loop/prompts/*.md` — editable policy for planner, builder, triage, coherence, and security roles.
+- `.github/workflows/loop-*.yml` — generated GitHub Actions mechanics.
+
+Edit the files under `.github/loop/` to steer the loop. Re-run
+`micro loop init --roles all --force` only when you want to regenerate workflow
+mechanics from the installed CLI.
+
+## 2. Configure the dispatch token
+
+The scheduled builder needs a repository secret containing a token from a user
+account that the coding agent will answer. Go Micro names that secret
+`CODEX_TRIGGER_TOKEN` by default. If you use another secret name, pass it when
+you initialize the loop:
+
+```bash
+micro loop init --agent @codex --token-secret LOOP_TOKEN --roles all
+```
+
+The token needs enough repository permission to open issues, comment, push
+branches, create pull requests, and enable auto-merge. Run `gh auth setup-git` in
+the environment that will push branches so `git push` uses the same credentials
+as `gh`.
+
+## 3. Make CI the gate
+
+The loop should not be its own reviewer. Protect the default branch so PRs merge
+only after the required checks pass. At minimum, require the same commands the
+Go Micro loop verifies locally and in CI:
+
+```bash
+go build ./...
+go test ./...
+golangci-lint run ./...
+```
+
+If your repository has a harness or end-to-end grader, make that required too.
+Keep human approval requirements out of the autonomous path unless you intend the
+loop to pause for review.
+
+## 4. Verify the wiring
+
+After editing the North Star, queue, prompts, token secret, and branch
+protection, run:
+
+```bash
+micro loop verify
+```
+
+`micro loop verify` checks that the loop direction, queue, prompts, role
+workflows, and non-loop CI gate are present. Fix any reported missing items
+before relying on scheduled increments.
+
+## 5. Operate the queue
+
+Keep one ranked list in `.github/loop/PRIORITIES.md`. Each item should link a
+scoped issue and be small enough for one PR. The builder closes both the priority
+issue and the per-run tracker issue in the PR body, for example:
+
+```text
+Closes #1234
+Closes #5678
+```
+
+Use the North Star to keep the queue honest: favor small improvements that move
+developers through the services → agents → workflows lifecycle, and surface
+breaking API or brand/positioning decisions for humans instead of auto-merging
+them.

internal/website/docs/guides/no-secret-first-agent.md

@@ -100,6 +100,7 @@ CI keeps those CLI boundaries present with:
 
 ```sh
 go test ./cmd/micro -run TestFirstAgentWalkthroughCLIBoundaries -count=1
+go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentDebuggingSmoke -count=1
 ```
 
 ## Debug transcript checkpoint

internal/website/docs/guides/your-first-agent.md

@@ -47,7 +47,7 @@ export ANTHROPIC_API_KEY=sk-ant-...
 Plain service calls work without a model key; the key is only needed when the
 agent reasons over tools.
 
-Run the read-only first-agent preflight before starting the walkthrough. The same CLI boundary is covered by CI with `go test ./cmd/micro -run TestFirstAgentWalkthroughCLIBoundaries -count=1`, so the documented scaffold → run → chat → inspect path stays visible in the local harness:
+Run the read-only first-agent preflight before starting the walkthrough. The same CLI boundary is covered by CI with `go test ./cmd/micro -run TestFirstAgentWalkthroughCLIBoundaries -count=1`, and the copy/paste tutorial code is built from a clean temporary workspace with `go test ./internal/harness/zero-to-hero-ci -run TestYourFirstAgentTutorialSmoke -count=1`, so the documented scaffold → run → chat → inspect path stays visible in the local harness:
 
 ```sh
 micro agent preflight
@@ -177,7 +177,14 @@ Create a task called "Review the first-agent walkthrough", then show me all task
 ```
 
 A healthy run shows the agent calling the task service and then summarizing the
-result. If the model refuses to call tools, tighten the prompt so it explicitly
+result. Inspect the recorded run when you want to see the tool calls, memory,
+and timing behind the answer:
+
+```sh
+micro inspect agent assistant
+```
+
+If the model refuses to call tools, tighten the prompt so it explicitly
 uses the `task` service before answering.
 
 ## 4. Know what just happened

internal/website/docs/guides/zero-to-hero.md

@@ -22,7 +22,7 @@ cloud credentials?"
 | First agent | `micro new`, `micro agent preflight`, `micro run`, `micro chat`, and `micro inspect agent <name>` stay available for the documented first-agent walkthrough. | `go test ./cmd/micro -run TestFirstAgentWalkthroughCLIBoundaries -count=1` |
 | Run | `micro run` remains the local development entry point. | `go test ./cmd/micro -run TestZeroToHeroCLIBoundaries -count=1` |
 | Chat | `micro chat` remains the interactive agent entry point. | `go test ./cmd/micro -run TestZeroToHeroCLIBoundaries -count=1` |
-| Inspect | `micro inspect agent <name>`, `micro inspect flow <flow>`, and `micro flow runs <flow>` remain discoverable for run history. | `go test ./cmd/micro -run TestZeroToHeroCLIBoundaries -count=1` |
+| Inspect | `micro inspect agent <name>`, `micro agent history <name>`, `micro inspect flow <flow>`, and `micro flow runs <flow>` remain discoverable for run history; the no-secret debugging smoke seeds durable agent history and runs the documented inspect/history commands without provider keys. | `go test ./internal/harness/zero-to-hero-ci -run TestNoSecretFirstAgentDebuggingSmoke -count=1` |
 | Deploy | `micro deploy --dry-run` resolves deploy targets without touching remote infrastructure. | `go test ./cmd/micro/cli/deploy -run TestDeployDryRun -count=1` |
 | Smallest first agent | `examples/first-agent` runs one service-backed agent with a deterministic mock model and no provider key. | `go test ./examples/first-agent -run TestRunFirstAgent -count=1` |
 | Runtime reference app | `examples/support` runs typed services, an agent using those services as tools, an event-driven flow handoff, and an approval gate with only the model mocked. | `go test ./examples/support -run 'TestRunSupportMockSmoke|TestZeroToHeroReadmeDocumentsLifecycle' -count=1` |

internal/website/docs/index.md

@@ -16,7 +16,7 @@ It's built on a pluggable architecture of Go interfaces: service discovery, clie
 
 ## Learn More
 
-Start with [Getting Started](getting-started.html) for install and the first local service. Then follow the first-agent on-ramp: `micro agent demo` for the installed no-secret CLI affordance, [No-secret first-agent transcript](guides/no-secret-first-agent.html) to run a mock-model support agent, [Your First Agent](guides/your-first-agent.html) to build and chat with a service-backed agent, [Debugging your agent](guides/debugging-agents.html) to inspect runs and memory, and the [0→hero reference path](guides/zero-to-hero.html) to walk the full scaffold → run → chat → inspect → deploy dry-run lifecycle covered by CI.
+Start with [Getting Started](getting-started.html) for install and the first local service. Then follow the first-agent on-ramp: `micro agent demo` for the installed no-secret CLI affordance, [examples wayfinding index](https://github.com/micro/go-micro/blob/master/examples/INDEX.md) for the maintained examples map, [the 0→hero support reference](https://github.com/micro/go-micro/tree/master/examples/support) for the full no-secret lifecycle example, [No-secret first-agent transcript](guides/no-secret-first-agent.html) to run a mock-model support agent, [Your First Agent](guides/your-first-agent.html) to build and chat with a service-backed agent, [Debugging your agent](guides/debugging-agents.html) to inspect runs and memory, and the [0→hero reference path](guides/zero-to-hero.html) to walk the full scaffold → run → chat → inspect → deploy dry-run lifecycle covered by CI.
 
 Otherwise continue to read the docs for more information about the framework.
 
@@ -25,10 +25,14 @@ Otherwise continue to read the docs for more information about the framework.
 - [Getting Started](getting-started.html)
 - [0→hero Reference](guides/zero-to-hero.html) - Walk scaffold → run → chat → inspect → deploy dry-run with CI-backed commands
 - `micro agent demo` - Show the provider-free first-agent demo command and next docs steps
+- `micro examples` - Show provider-free first-agent examples in copy/paste order
+- [Examples wayfinding index](https://github.com/micro/go-micro/blob/master/examples/INDEX.md) - Choose the first-agent, support, and interop examples from one map
+- [0→hero support reference](https://github.com/micro/go-micro/tree/master/examples/support) - Run the maintained no-secret services → agents → workflows example
 - [No-secret first-agent transcript](guides/no-secret-first-agent.html) - Run the first useful agent path without a provider key
 - [Your First Agent](guides/your-first-agent.html) - Build a service-backed agent end to end
 - [MCP & AI Agents](mcp.html) - Turn services into AI-callable tools with the Model Context Protocol
 - [CLI & Gateway Guide](guides/cli-gateway.html) - Development vs Production modes
+- [`micro loop` quickstart](guides/micro-loop.html) - Scaffold an autonomous CI-gated improvement loop
 - [Quick Start](quickstart.html)
 - [Architecture](architecture.html)
 - [Configuration](config.html)
@@ -63,5 +67,6 @@ Otherwise continue to read the docs for more information about the framework.
 - [Real-World Examples](examples/realworld/)
 - [Migration Guides](guides/migration/)
 - [Observability](observability.html)
+- [`micro loop` quickstart](guides/micro-loop.html)
 - [Contributing](contributing.html)
 - [Roadmap](roadmap.html)

internal/website/docs/quickstart.md

@@ -43,12 +43,14 @@ You now have the service half of the services → agents → workflows lifecycle
 
 1. **[Install troubleshooting](guides/install-troubleshooting.html)** - verify the binary installer or `go install`, `PATH`, `micro --version`, and the no-secret smoke path.
 2. `micro agent demo` - print the provider-free first-agent demo command and the next docs steps from the installed CLI.
-3. `micro zero-to-hero` - print the maintained one-command no-secret lifecycle harness and runnable examples.
-4. **[Smallest first-agent example](https://github.com/micro/go-micro/tree/master/examples/first-agent)** - run a mock-model, no-secret agent before adding provider keys.
-5. **[No-secret first-agent transcript](guides/no-secret-first-agent.html)** - run a useful support agent with a mock model before setting up a provider key.
-6. **[Your First Agent](guides/your-first-agent.html)** - turn this service into an agent-callable tool, chat with it, and learn the `micro agent preflight` → `micro run` → `micro chat` loop.
-7. **[Debugging your agent](guides/debugging-agents.html)** - inspect service registration, tool calls, run history, memory, provider failures, and flow handoffs when the agent does something surprising.
-8. **[0→hero Reference](guides/zero-to-hero.html)** - walk the maintained scaffold → run → chat → inspect → deploy dry-run path that proves services, agents, and workflows together.
+3. `micro examples` - print the maintained provider-free runnable examples in copy/paste order.
+4. `micro zero-to-hero` - print the maintained one-command no-secret lifecycle harness and runnable examples.
+5. **[Examples wayfinding index](https://github.com/micro/go-micro/blob/master/examples/INDEX.md)** - choose the smallest no-secret first-agent, maintained **[0→hero support reference](https://github.com/micro/go-micro/tree/master/examples/support)**, and next interop examples from one map.
+6. **[Smallest first-agent example](https://github.com/micro/go-micro/tree/master/examples/first-agent)** - run a mock-model, no-secret agent before adding provider keys.
+7. **[No-secret first-agent transcript](guides/no-secret-first-agent.html)** - run a useful support agent with a mock model before setting up a provider key.
+8. **[Your First Agent](guides/your-first-agent.html)** - turn this service into an agent-callable tool, chat with it, and learn the `micro agent preflight` → `micro run` → `micro chat` loop.
+9. **[Debugging your agent](guides/debugging-agents.html)** - inspect service registration, tool calls, run history, memory, provider failures, and flow handoffs when the agent does something surprising.
+10. **[0→hero Reference](guides/zero-to-hero.html)** - walk the maintained scaffold → run → chat → inspect → deploy dry-run path that proves services, agents, and workflows together.
 
 After that first-agent path, branch out to:
 

micro.go

@@ -220,6 +220,13 @@ func AgentResume(ctx context.Context, a Agent, runID string) (*AgentResponse, er
 	return agent.Resume(ctx, a, runID)
 }
 
+// AgentResumePending resumes every incomplete checkpointed agent run, oldest
+// first. It returns the first run id that fails again so startup recovery loops
+// can leave the durable backlog visible instead of swallowing the failure.
+func AgentResumePending(ctx context.Context, a Agent) (string, error) {
+	return agent.ResumePending(ctx, a)
+}
+
 // AgentResumeInput resumes a checkpointed agent run waiting for human input.
 func AgentResumeInput(ctx context.Context, a Agent, runID, input string) (*AgentResponse, error) {
 	return agent.ResumeInput(ctx, a, runID, input)