AndroidAgenticWiki

A Git-backed knowledge graph and Model Context Protocol (MCP) server for Android development.

Standard LLMs lack real-time awareness of the rapidly evolving Android ecosystem. AndroidAgenticWiki bridges this gap by serving as a "Long-Term Memory" vault. By connecting this repository to your IDE via MCP, your AI agent gains access to local constraints that override general training data with workspace-specific rules and official Google standards, eliminating hallucinations of deprecated APIs.

📦 What's Included (Out-of-the-Box)

You don't just get an empty folder structure. This vault comes pre-loaded with Senior-level architectural constraints:

• Build Rules (build_constraints.md): Enforces AGP 9.2.0+, Compose BOM 2026.04.01, and strict declarative setups (banning the legacy kotlin-android plugin).
• UI Standards (ui_standards.md): Mandates stable Material 3, Edge-to-Edge enforcement, and Compose Modifier.graphicsLayer performance rules.
• Architecture (architecture.md): Teaches the agent to auto-detect and enforce strict Clean Architecture, MVVM, or MVI patterns based on your project structure.
• Sanitized Blueprints: Actionable templates for koin_di.md, mvvm_state.md, room3_configuration.md, and design_md_implementation.md.
• Google Android Skills: Automatically syncs with the official android/skills repository via Git submodules for authoritative navigation, performance, and UI guidance.

🧠 The Architecture

This brain utilizes two components operating on the exact same local folder simultaneously:

1. The MCP Server (Backend): Acts as the headless bridge into your Android Studio AI. You can run this via Tolaria (a dedicated desktop app) or natively via Node.js.
2. Obsidian (GUI & Data Funnel): Acts as your optional visualizer. You use it to view the knowledge graph and clip new Android documentation from the web directly into the vault.

⚠️ CRITICAL WARNING: DO NOT DOWNLOAD THE ZIP FILE This repository relies on Git Submodules to synchronize the official Google Android Skills. If you use GitHub's "Download ZIP" feature, the external_skills/ directory will be completely empty and the AI agent will fail to route correctly. You must use the npx command or install via Git clone.

🚀 Setup Guide

1. Initialize the Vault

The easiest way (One-Click Setup via npx): Simply run the following command in your terminal. It will automatically download the repository, initialize the Google skills submodules, and auto-configure your AI assistants (Claude, Gemini CLI, Codex):

npx android-agentic-wiki@latest my-android-brain

Alternative (Manual Git Clone): If you prefer to clone manually:

git clone https://github.com/trinadhthatakula/AndroidAgenticWiki.git
cd AndroidAgenticWiki
git submodule update --init --recursive
npm install
npm run setup

2. Choose Your MCP Server Path

Option A: Tolaria (Recommended for macOS)

1. Download and install Tolaria.
2. Open this repository folder as a Vault in Tolaria.
3. Open Tolaria's Command Palette and search "Set Up External AI Tools" to copy your unique MCP Server Command path.

Option B: Native Node.js (Windows / Linux / CLI Preference) This repository acts as its own MCP server.

1. Ensure Node.js (v18+) is installed on your machine.
2. Open a terminal in the root of the cloned repository and run npm install.
3. To test the server, run npm run mcp. (You can stop it with Ctrl+C after verifying it boots).

3. IDE Integration & Configuration

You can wire the knowledge graph into your AI assistant using either our auto-setup script, UI configuration, or manual configuration depending on your client.

Method A: Auto-Setup (Claude Desktop, Claude Code, Gemini CLI, Codex) If you installed the vault using npx create-android-agentic-wiki, your IDEs are already configured automatically.

If you cloned the repository manually, you can run the setup script in the root of the cloned repository:

npm run setup

(This automatically configures claude_desktop_config.json, ~/.claude.json, ~/.gemini/settings.json, and ~/.codex/config.toml).

Method B: UI Configuration (Android Studio / IntelliJ)

1. Open Settings (Windows/Linux) or Preferences (macOS).
2. Navigate to Plugins > AI > MCP Servers.
3. Click Add Server (+).
4. Name: AndroidAgenticWiki
5. Executable / Command: npx
6. Arguments: -y @modelcontextprotocol/server-filesystem /absolute/path/to/your/AndroidAgenticWiki (Crucial: Replace the path with the actual absolute path to the cloned repository).
7. Restart the AI Assistant.

Method C: Manual Configuration (Cursor, Custom Clients) If your AI client uses a JSON configuration file (e.g., claude_desktop_config.json), copy and paste the following block into your mcpServers object.

Remember to replace /absolute/path/to/your/AndroidAgenticWiki with your actual local path.

{
  "mcpServers": {
    "android-agentic-wiki": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/absolute/path/to/your/AndroidAgenticWiki"
      ]
    }
  }
}

For Codex (~/.codex/config.toml), use:

[mcp_servers."android-agentic-wiki"]
command = "npx"
args = ["-y", "@modelcontextprotocol/server-filesystem", "/absolute/path/to/your/AndroidAgenticWiki"]

4. Configure the Data Funnel (Obsidian)

1. Download Obsidian and open the AndroidAgenticWiki folder as a vault.
2. Install the Obsidian Web Clipper browser extension.
3. In the Web Clipper settings, go to Templates > Default.
4. Set the Folder location to inbox.
5. Add a new Property: Name = status, Value = unprocessed.

🤖 Your First Prompt

Once connected, open the AI chat in Android Studio and trigger the routing engine:

"Based on the AGENTS.md routing in my AndroidAgenticWiki, create a new User Profile feature using the MVVM blueprint and Koin DI constraints."

📥 How the Wiki Learns (The Ingestion Workflow)

This repository is not a static template; it is a self-updating system. You can literally "teach" your agent new Android concepts, libraries, or architectural patterns without writing markdown yourself.

When a new Android feature, library update, or architecture blog post is released:

1. Clip the Knowledge: Use the Obsidian Web Clipper to save the raw Medium article or Android Developer doc directly into your /inbox folder.
2. Trigger the Brain: Open Android Studio and tell your connected agent: "Execute the runbook in workflows/ingest_docs.md on the unprocessed files in my inbox."
3. Watch it Learn: The agent will autonomously read the article, extract the strict engineering rules, update your build_constraints.md or ui_standards.md, write a fresh code blueprint in /blueprints/, and archive the raw document.

Result: The next time you ask the agent to build a feature, it will natively use the new architecture it just learned.

⚠️ Disclaimer

The rules and blueprints provided in this repository are highly opinionated templates designed for modern, high-performance Android development. Every team is different. You are highly encouraged to fork this repository and modify the /rules/ and /blueprints/ directories to perfectly match your own tech stack and organizational standards.