# AppMap

AppMap helps developers and AI agents understand what AI-written code actually does at runtime. It builds a behavioral model of how your software runs, in your editor, so every change can be reviewed and trusted. The model can travel with the pull request and is read by any AI agent over MCP. Free, open source, local.

AppMap Gold Traces are recordings promoted into approved before/after behavioral baselines: invariant flows must hold across a change, intended changes must show up. Deterministic traces compare byte-for-byte; real-world traces compare normalized behavior.

## Why it matters for AI agents

- Runtime context: Coding agents often work without the application environment, database, or credentials. Source code can show possible execution paths, but not which calls, queries, and side effects occurred in a specific run. AppMap traces supply those runtime facts with the repository.

- Independent evidence: AppMap records traces from the running application, outside the agent's reasoning. Agents can query the traces, but they do not create them.

## Where to look
- Home: https://appmap.io/
- How it works: https://appmap.io/how-it-works
- Gold Traces: https://appmap.io/gold-traces
- Architecture: https://appmap.io/architecture
- Benchmarks: https://appmap.io/benchmarks
- Compatibility: https://appmap.io/how-it-works#compatibility
- Enterprise: https://appmap.io/enterprise
- Research paper (read online): https://appmap.io/research/runtime-rca
- GitHub: https://github.com/getappmap