Skip to content

agentplay

agentplay: replay any coding agent session

agentplay replays your coding agent sessions as a map and a timeline: what the agent read, what it edited, what failed and what it shipped. It runs locally with one command, and it is open source under Apache 2.0.

Share on XRSS

One command, every session on your machine#

Run uvx agentplay@latest. agentplay finds the sessions your agents already wrote to disk, builds a redacted replay of each one and opens a local viewer. It works across Claude Code, Codex, Cursor, Copilot, OpenCode, Kiro, Gemini CLI, Antigravity CLI and Command Code.

A map of everything the session touched#

Each session becomes a radial map with the conversation at the centre: repositories and files coloured by whether they were searched, read or edited; MCP servers and the tools called through them; every model that served a request; subagents with their own calls and cost; and errors on the outermost branch.

A timeline you can play back#

Under the map, the replay plays the session in order, so you can see what the agent read before it edited a file, which commands it retried, where the context window filled up and which PRs, commits and tickets it produced.

Local, and open source#

It runs without an account or an API key, and session data stays on your machine. agentplay has no runtime dependencies beyond the Python standard library, and the code is on GitHub under Apache 2.0. For the same view across a whole team, see Agent Insights.

Improvements#

  • agentplayRuns with uv, pipx or pip on Python 3.11 or newer.
  • agentplayShips with a bundled demo session built from synthetic data.
  • agentplaySet AGENTPLAY_NO_NETWORK=1 to switch off the one optional network call.

The memory starts the day you connect a repo.

From that day on, every session, review and incident lands in the same index.

  • No credit card required
  • No training on customer code
  • Deploy in your cloud, your own keys