Claude-replay – Replay your Claude Code sessions

Name: Claude-replay – Replay your Claude Code sessions
Availability: InStock
Author: kouiskas

by kouiskas·Mar 2, 2026·5 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidCozyNiche Gem

Session replay TUI for Claude Code learning, but limited to existing session logs.

Strengths

•Elegant Bubble Tea UI with rich filtering, navigation, and multiple export formats (asciinema, GIF, MP4) for demos and documentation.
•Git branch integration (`claude-session-trail` companion) enables team knowledge sharing without painful manual exports.
•Addresses real friction: reviewing why an agent made a decision, debugging failed runs, or showing others how Claude Code solved a problem.

Weaknesses

•Only reads existing Claude Code session logs—doesn't capture broader agent reasoning or integrate with other AI tools.
•Niche to Claude Code ecosystem; limited utility if your team uses multiple agent frameworks.

Post Description

I've been helping teams learn how to leverage Claude Code the best and I've built a couple of complimentary tools to help people learn from each other.

This replay TUI, claude-replay, which lets you explore and revisit your past Claude Code sessions. And a companion optional Claude Code plugin, claude-session-trail, to automatically store your sessions in a dedicated git branch: https://github.com/Trailblaze-work/claude-session-trail

Both tools can work in tandem, as the replay client can also read project sessions from the git branch, in addition to its default of accessing the user's local sessions.

This is hot off the press, so there might be some rough edges still, but the base functionality you would expect works. I think they have purpose as learning tools at a minimum, and can be useful for self-introspection.

The automatic committing is probably too much for mature teams, for whom the export/sharing functionalities of specific sessions might be enough.

Yes, models change and what works in terms of prompting strategy now might not work on newer models in the future, but at the moment everyone develops strategies and good habits locally in parallel. Being able to look over someone's shoulders and seeing how they do it might be useful. And it lowers the effort required to share how you work inside Claude Code with others if you wish to do so.

Contributions welcome!