# arcade-agent
> Open-source (MIT) architecture analysis for coding agents: recovers
> components/dependencies/smells from code, serves them over MCP (**18 tools**),
> tracks drift via a git-committed baseline.

## Docs
- Getting started: https://arcade-agent.dev/docs/getting-started/
- Agent tools (MCP setup + all tools): https://arcade-agent.dev/docs/agent-tools/
- Star tools: context_for_task, diff_impact, api_surface, dependency_cone — https://arcade-agent.dev/docs/agent-tools/context-for-task/, https://arcade-agent.dev/docs/agent-tools/diff-impact/, https://arcade-agent.dev/docs/agent-tools/api-surface/, https://arcade-agent.dev/docs/agent-tools/dependency-cone/
- CI drift baseline: https://arcade-agent.dev/docs/ci/
- Research lineage (ARCADE workbench, algorithm provenance, publications): https://arcade-agent.dev/research/
- Self x-ray (the tool's own arcade-self-analysis report, published unmodified): https://arcade-agent.dev/self/
- Concepts (newcomer architecture primer): https://arcade-agent.dev/docs/concepts/what-is-software-architecture/, https://arcade-agent.dev/docs/concepts/components-and-dependencies/, https://arcade-agent.dev/docs/concepts/the-c4-model/, https://arcade-agent.dev/docs/concepts/architecture-recovery/, https://arcade-agent.dev/docs/concepts/smells-and-metrics/, https://arcade-agent.dev/docs/concepts/drift/
- Metrics reference (formulas as implemented, self-analysis worked examples): https://arcade-agent.dev/docs/reference/metrics/

## Source
- GitHub: https://github.com/lemduc/arcade-agent
- PyPI: https://pypi.org/project/arcade-agent/