# Mirrors

> Replay your AI agent's real production sessions against your next change and catch broken agents before your customers do.

## Docs

- [Collect traces](https://runmirrors.com/docs/api-reference/collect.md): POST /api/collect — the trace ingest endpoint the collectors speak.
- [Environments](https://runmirrors.com/docs/api-reference/envs.md): List, inspect, and query your mirrors over /v1.
- [Introduction](https://runmirrors.com/docs/api-reference/introduction.md): The programmatic surface: workspace keys, versioning, and limits.
- [CI gate](https://runmirrors.com/docs/ci-gate.md): Block agent regressions at the pull request, Vercel-style.
- [CLI](https://runmirrors.com/docs/cli.md): The mirrors command — full parity with the web app, from your terminal.
- [Go](https://runmirrors.com/docs/collectors/go.md): The trace collector for Go agents — you build the trace, it ships it.
- [Python](https://runmirrors.com/docs/collectors/python.md): The two-line drop-in trace collector for Python agents.
- [TypeScript / JavaScript](https://runmirrors.com/docs/collectors/typescript.md): The drop-in trace collector for Node.js agents.
- [How it works](https://runmirrors.com/docs/how-it-works.md): Traces in, a runnable copy of your agent's world out.
- [What is Mirrors?](https://runmirrors.com/docs/index.md): Catch broken agents before your customers do.
- [MCP server](https://runmirrors.com/docs/mcp-server.md): Drive Mirrors from any AI client — hosted, OAuth, nothing to install.
- [Quickstart](https://runmirrors.com/docs/quickstart.md): Connect your agent, build a mirror, and gate merges in CI.