Local GGUF chat inference through llama.cpp.
Cloud chat engines through OpenAI-compatible and Anthropic Messages protocols.
User-configured web search with source cards and per-turn context injection.
Local and cloud image-generation engine management.
ModelScope-oriented model discovery and resumable downloads.
A Compose mobile UI for chat, model management, image generation, agent diagnostics, settings, and local API tools.

MCA does not include model weights or API keys. Users bring their own local models, cloud endpoints, and provider credentials.

Screenshots

Real-device screenshots from the Android app:

Chat	Workspace	Images	Model recommendations

View more screenshots

Settings	Cloud engines	Local engines	Model market

Model picker	Local API

The Local API screenshot is redacted before publishing.

The lightweight GIF above is generated from real-device screenshots. A higher quality MP4 is available at docs/assets/demo/mca-demo.mp4.

Status

This repository is an active Android app workspace. The chat and model management surfaces are usable, while local image generation is still experimental and should be tested per device and model bundle.

Current release status:

Alpha APKs are published through GitHub Releases.
The first public package target is arm64-v8a Android devices.
Local chat is the primary stable local path.
Web search is available after the user configures a search provider in Settings. MCA supports manual, smart-auto, and always-on trigger modes for SearxNG, Brave Search, Tavily, Jina Search, and custom JSON search endpoints.
Local image generation is experimental and requires complete model bundles. Do not treat phone-side image generation as a guaranteed stable feature yet.

Features

Local chat: native llama.cpp bridge, streaming generation, stop support, token speed labels, reasoning-content filtering, and local benchmark support.
Cloud chat: user-configured OpenAI-compatible or Anthropic Messages endpoints with locally encrypted API key storage.
Smart web search: user-configured SearxNG, Brave Search, Tavily, Jina Search, or custom JSON search providers. MCA can identify URLs, detect explicit search intent, expand time-sensitive or documentation-style queries, rank/deduplicate sources, summarize results into the current turn only, and display source cards under the assistant reply. Trigger modes are manual, smart-auto, and always-on; the Settings page keeps recent local search diagnostics for trigger reasons, closed-loop evidence, provider errors, partial expanded-query warnings, expanded queries, source counts, latency, clickable source URLs, provider labels, source snippets, and a local source quality score based on usable sources, readable content length, independent hosts, and safety blocks. Direct URL reading works after web search is enabled even before a search API is configured; keyword search still requires SearxNG, Brave, Tavily, Jina, or a custom JSON endpoint. Custom JSON endpoints can use URL templates such as /search?q={query}&limit={max_results} or common q/query/max_results parameters. Provider endpoints may be self-hosted, but readable page fetching and direct URL reading block localhost, private LAN, link-local, and reserved addresses by default for safety. When Jina Search is selected with a key, MCA can fall back to Jina Reader for public pages whose direct readable content is too weak, while keeping the same private-network guard. MCA reads multiple direct URLs, expanded search queries, and fetched page bodies concurrently to keep live search responsive on mobile networks. Keyword search successes use a short in-memory local cache to avoid repeating the same provider call within a brief window; direct URL reads are not cached and API keys are never stored in cache entries. The search test in Settings uses the fields currently typed into the form, so users can verify an endpoint before saving it. The closed-loop self-test records whether MCA produced provider results, prompt context, source-card data, quality scoring, and local diagnostics. Settings also includes a no-key public JSON self-check filler so users can verify the integration path before entering their own provider. MCA labels this source as 公开 JSON 自检源 in the app because it is only a protocol check with limited coverage, not a general web-search engine; production use should rely on a trusted or self-hosted search service. Custom JSON endpoints may return a top-level array, an object containing results, items, data, hits, or organic_results, or nested variants such as data.results and response.items. It accepts common URL/title fields such as url, link, href, html_url, story_url, canonical_url, displayLink, formattedUrl, source.url, title, full_name, story_title, and source.title, plus summary/body fields such as summary, excerpt, and pageContent. When smart query expansion creates multiple searches, MCA keeps successful sources even if one expanded query fails. Tavily and Jina use Authorization: Bearer <key>; Brave uses X-Subscription-Token and supports both the Web Search endpoint and the LLM Context endpoint for AI grounding/RAG-style snippets. Public SearxNG instances often rate-limit or disable JSON responses, so a self-hosted or explicitly approved endpoint is recommended for reliable search. Brave and Tavily official API root URLs are accepted and normalized to their search paths during preflight and request execution.
Image page: MCA image workspace with local/cloud engine switching, prompt composer, generation states, template cards, and image library.
Local image engines: stable-diffusion.cpp bridge with progress/cancel hooks and bundle-aware model registration.
Model hub: local/imported models, ModelScope recommendations, resumable downloads, file classification, and engine grouping.
Assistants and role cards: multiple local assistants with system prompts, default model preference, generation parameters, memory/search toggles, and JSON role-card import/export. MCA exports its own mca.assistant.card schema and can import common nested character-card data fields into a usable system prompt.
Agent diagnostics: local device profiling, model recommendations, benchmark-based tuning, and explainable parameter plans.
Local API: OpenAI-compatible local server for trusted same-device and same-LAN clients, including /v1/models, /v1/chat/completions, JSON replies, and SSE streaming.

Install

Download the latest alpha APK from GitHub Releases. Android may ask you to allow installation from your browser or file manager.

The APK does not include model weights or cloud credentials. After installing:

Add a local GGUF chat model or configure a cloud chat engine.
Configure an image engine if you want cloud or local image generation.
Configure web search in Settings if you want live search. You can use a self-hosted SearxNG endpoint, Brave Search, Tavily, Jina Search, or a compatible custom JSON endpoint, test the current form values, then choose manual, smart-auto, or always-on triggering. Brave can use either /res/v1/web/search for normal search or /res/v1/llm/context for grounding-oriented snippets; Brave/Tavily official root URLs are auto-filled to the normal search paths. Direct page reading will refuse localhost, LAN, link-local, and reserved addresses unless a development build explicitly enables private-network fetching. See docs/WEB_SEARCH.md for the full configuration, trigger-mode, source-card, and troubleshooting guide. Quick source guide: choose Tavily or Brave for the fastest API-key setup, choose self-hosted SearxNG when privacy and control matter most, choose Jina when page-body extraction needs help, or choose custom JSON when you operate your own search gateway.
Check docs/PERMISSIONS.md before enabling network or local API workflows.
Check docs/MODEL_COMPATIBILITY.md before choosing local image bundles or cloud provider protocols.

Release APKs are signed by the project maintainer. Debug APKs are not intended for public installation.

Optional live web-search smoke tests for maintainers:

$env:MCA_LIVE_WEB_SEARCH_TEST='true'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveDirectUrlSmokeReadsRealWebPageWhenEnabled

$env:MCA_LIVE_SEARXNG_ENDPOINT='https://your-searxng.example'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveSearxngSmokeUsesConfiguredEndpointWhenProvided

$env:MCA_LIVE_BRAVE_API_KEY='<key>'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveBraveSmokeUsesConfiguredKeyWhenProvided

$env:MCA_LIVE_TAVILY_API_KEY='<key>'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveTavilySmokeUsesConfiguredKeyWhenProvided

$env:MCA_LIVE_JINA_API_KEY='<key>'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveJinaSmokeUsesConfiguredKeyWhenProvided

$env:MCA_LIVE_CUSTOM_JSON_ENDPOINT='https://hn.algolia.com/api/v1/search'
.\gradlew :app:testDebugUnitTest --tests com.muyuchat.mca.WebSearchProviderTest.liveCustomJsonClosedLoopBuildsPromptSourcesAndDiagnosticsWhenProvided

Repository Layout

:app - Android application, navigation, ViewModel, cloud/local providers.
:core:native - llama.cpp C++/JNI bridge for local chat.
:core:sd-native - stable-diffusion.cpp C++/JNI bridge for local image generation.
:core:engine - single-active-generation inference service.
:core:modelstore - GGUF import, manifest, SHA-256, managed model storage.
:core:download - ModelScope parsing, file listing, and resumable downloads.
:core:telemetry - runtime metrics, SoC detection, and JSONL logging.
:core:deviceprofile - device capability and thermal profiling.
:core:tuning - parameter plan generation.
:core:benchmark - short local benchmark runner.
:core:advisor - local recommendation engine and agent logs.
:api:local - AIDL service and loopback REST server skeleton.
:feature:chat - chat and image-generation UI.
:feature:agent - agent diagnostics UI.
:feature:modelhub - model management UI.
:feature:settings - runtime, logs, and local API UI.

Build Requirements

Android Studio or command-line Gradle.
JDK 17.
Android SDK with:
- Android platform matching compileSdk.
- Android build tools.
- CMake 3.31.6 or compatible version configured by Gradle.
- Android NDK matching gradle/libs.versions.toml.

Create a local local.properties file or set ANDROID_HOME:

sdk.dir=/path/to/android-sdk

local.properties is ignored by Git.

Clone

This repository uses submodules for native inference backends:

git clone --recurse-submodules <repo-url>
cd mym

If you cloned without submodules:

git submodule update --init --recursive

Build

PowerShell:

$env:JAVA_HOME='<path-to-jdk-17>'
$env:ANDROID_HOME='<path-to-android-sdk>'
.\gradlew.bat :app:assembleDebug

Bash:

export JAVA_HOME=/path/to/jdk-17
export ANDROID_HOME=/path/to/android-sdk
./gradlew :app:assembleDebug

The debug APK is generated under:

app/build/outputs/apk/debug/

Native Backends

Local Chat

core/native builds libmca_native.so. With third_party/llama.cpp present, the module links the llama.cpp Android CPU backend. The project keeps a stub fallback for development builds where llama.cpp is temporarily unavailable.

Local Image Generation

core/sd-native builds libmca_sd_native.so against third_party/stable-diffusion.cpp. MCA stores its Android-specific patch in:

third_party/patches/stable-diffusion.cpp-mca-android.patch

Gradle applies this patch when needed before the native CMake build.

Local image generation is model-bundle sensitive. Some newer image models need a diffusion model plus VAE/AE and text encoder/LLM components in the same engine directory. It is currently an experimental capability and should be validated on each target device before being promoted as stable.

Model and API Compatibility

See docs/MODEL_COMPATIBILITY.md for the current compatibility matrix covering local GGUF chat, OpenAI-compatible chat, Anthropic Messages, OpenAI Images, DashScope Image, custom image paths, and experimental local image bundles.

Local API

MCA can expose the currently loaded local chat model through an OpenAI-compatible API for trusted clients.

Use this when you want another app, browser, desktop client, or local tool to talk to the model running on the phone.

Recommended client settings:

Field	Value
Protocol	OpenAI-compatible
Same-device Base URL	`http://127.0.0.1:11435/v1`
Same-LAN Base URL	`http://<phone-lan-ip>:11435/v1`
API key	Generated inside MCA Settings -> Local API
Model	Pick from `/v1/models`, or enter the returned model `id` manually

Supported paths:

GET /health
GET /v1/models
POST /v1/chat/completions
GET / for the built-in web chat page

/v1/chat/completions supports standard JSON responses and stream=true Server-Sent Events. Same-LAN access requires enabling the in-app "open port" switch and should only be used on trusted networks.

Privacy

Local chat and local image generation run on device.
Cloud chat and cloud image generation send prompts to the user-configured provider endpoint.
Cloud API keys are stored locally with Android Keystore-backed encryption.
The repository does not contain API keys, model weights, or private user data.

See PRIVACY.md and docs/PERMISSIONS.md for more detail.

Roadmap

v0.1 alpha: local chat, cloud chat, cloud image engines, image workspace, model management, ModelScope-oriented downloads, and release packaging.
v0.2 alpha: smart web search, source cards, role-card assistants, local API compatibility fixes, web-search diagnostics, and release-grade compatibility documentation.
v0.3: stabilize local image bundles, improve device compatibility reporting, and refine image generation progress/cancel behavior.

MCA intentionally avoids bundling model weights. Model recommendations and download sources must respect each upstream model's license.

Third-Party Code

The repository references upstream native projects through submodules:

llama.cpp - MIT License.
stable-diffusion.cpp - MIT License.

See THIRD_PARTY_NOTICES.md.

Contributing

MCA is early-stage and changes quickly. Please read CONTRIBUTING.md before opening issues or pull requests.

License

MCA is licensed under the MIT License. See LICENSE.