e071084ebe
govulncheck / govulncheck (push) Waiting to run
Harness (E2E) / Harnesses (mock LLM) (push) Waiting to run
Harness (E2E) / Provider harnesses (live LLM conformance) (push) Waiting to run
Lint / golangci-lint (push) Waiting to run
Run Tests / Unit Tests (push) Waiting to run
Run Tests / Etcd Integration Tests (push) Waiting to run
Agent Demo
A multi-service project management app that demonstrates AI agents interacting with Go Micro services through MCP.
What's Included
Three services registered in a single process:
| Service | Endpoints | Description |
|---|---|---|
| ProjectService | Create, Get, List | Manage projects with status tracking |
| TaskService | Create, List, Update | Tasks with assignees, priorities, and status |
| TeamService | Add, List, Get | Team members with roles and skills |
The demo starts with seed data: 2 projects, 7 tasks, and 4 team members.
Run
go run main.go
Endpoints:
- MCP Gateway: http://localhost:3000
- MCP Tools: http://localhost:3000/mcp/tools
- WebSocket: ws://localhost:3000/mcp/ws
Use with Claude Code
{
"mcpServers": {
"demo": {
"command": "go",
"args": ["run", "main.go"],
"cwd": "examples/agent-demo"
}
}
}
Example Prompts
Try these with Claude Code or any MCP client:
- "What projects do we have?"
- "Show me all tasks assigned to alice"
- "Create a high-priority task for bob to review the design mockups"
- "Who on the team knows Go?"
- "Give me a status update on the Website Redesign project"
- "What tasks are still todo on the API v2 migration?"
- "Assign the unassigned tasks to charlie"
- "Mark task-1 as done"
What This Demonstrates
- Zero-config MCP — Services become AI tools automatically from doc comments
- Cross-service orchestration — An agent queries projects, tasks, and team in one conversation
- Rich tool descriptions —
descriptionstruct tags and@examplecomments guide the agent - Auth scopes — Read and write operations have separate scopes
WithMCPone-liner — MCP gateway starts with a single option
See the blog post for a detailed walkthrough.