A few weeks ago, I posted a CLI version of this — thanks for the great feedback, it's been going well (500+ stars).

I originally built it for demos. But people have been using it for knowledge sharing, teaching, documentation, and even debugging sandboxed agent workflows in the cloud. Pretty cool to learn about these use cases.

Since then: Cursor and Codex CLI support, Docker, --watch live mode, custom redaction, and two new ways to use it:

- Web UI, drag and drop a session file, no install needed: https://es617.dev/claude-replay/

- Local editor, run claude-replay to launch a full editor in your browser: browse sessions, edit prompts, exclude turns, add bookmarks, live preview

The output is still a single self-contained HTML file — step through prompts, tool calls, and reasoning. Zero dependencies, works on mobile, embeddable via iframe.

Repo: https://github.com/es617/claude-replay

  • vanyaland 3 hours ago |
    This is a good direction. With longer coding sessions, the final diff is only part of the story. The hard part is making the replay readable enough that we can spot the important moments quickly instead of scrubbing through noise.