agentplay

Replay your coding agent sessions as a map and a timeline

About

See what your coding agent actually did, step by step

Run `uvx agentplay` and it finds the agent sessions already in ~/.claude, ~/.codex, ~/.cursor and similar folders, builds a redacted replay of each one and opens a local web viewer. Each session shows as a radial map of files, tools, MCP servers, subagents and models, with a replay timeline underneath. It helps answer questions like which files the agent read before editing, which commands failed and how often it retried them, which MCP servers errored, where the context window filled up and what subagents brought back. Claude Code sessions get the most detail (hooks, compactions, stop reasons, latency, subagent lanes). No account, no API key and no runtime dependencies beyond the Python standard library; session data never leaves your machine, and AGENTPLAY_NO_NETWORK=1 turns off the optional email signup in the viewer.

Features

Reads Claude Code, Codex, Cursor, Copilot CLI, Copilot in VS Code, OpenCode and Kiro sessions
Radial map of files, tools, MCP servers, subagents and models per session
Playable timeline of every step
Spots failed commands, retries and context-window pressure
Redacts secrets before building replays
Local only, no account or API key

Install