MCP
Arxo MCP
Section titled “Arxo MCP”The Arxo MCP (Model Context Protocol) server exposes architecture metrics, cycle detection, LLM integration health, policy evaluation, and refactoring suggestions to AI assistants such as Cursor and Claude. Use it to run analyses, check policies, and get impact and refactoring suggestions from your IDE or chat.
Supported Languages
Section titled “Supported Languages”The MCP server analyzes the same languages as the Arxo engine: TypeScript, JavaScript, Rust, Python, Java, Kotlin, and Go. Create an arxo.yaml in your project root (optional) to configure analysis; see Configuration for config options.
When to Use MCP vs CLI vs IDE
Section titled “When to Use MCP vs CLI vs IDE”| Use case | Best option |
|---|---|
| Chat-driven analysis (“Check cycles in this project”) | MCP — AI calls tools on demand |
| Pre-commit hooks, CI/CD pipelines | CLI — arxo analyze, arxo doctor |
| Real-time analysis as you code, inline violations | IDE extension — Arxo VS Code extension |
MCP is ideal when you want to ask natural-language questions and get architecture insights within the AI chat, without leaving your editor.
Features
Section titled “Features”- Architecture analysis — Full analysis or fast cycle checks via MCP tools
- Policy evaluation — Evaluate policy invariants and get violations
- LLM integration health — Check observability gaps, PII risks, cost tracking
- Impact analysis — Blast radius and affected modules for specific files
- Refactoring suggestions — Recommendations from Ricci curvature and hotspots
- Resources — Metrics, violations, and dependency graphs via MCP resources
How it works
Section titled “How it works”You build the arxo-mcp binary and configure your MCP client (e.g. Cursor) to run it. The server communicates over stdio using the Model Context Protocol. AI assistants can then call tools (e.g. check_cycles, analyze_architecture) and read resources (e.g. current metrics, policy violations) for the project in the workspace.
Next Steps
Section titled “Next Steps”- Quick Start — Get running in under 5 minutes
- Installation — Build the binary and verify
- Configuration — Cursor and other MCP client setup
- Tools — All available MCP tools
- Resources — URI patterns and resource types
- Workflows — Example AI assistant workflows, caching, logging