SessionPatch vs claude-replay
August 2026
Both tools read Claude Code session files from ~/.claude/projects/. They solve different
problems, and plenty of people want both.
| SessionPatch | claude-replay | |
|---|---|---|
| Core job | Inspect, edit supported content, and diagnose session structure | Export a session as a shareable HTML replay |
| Runs | In the browser, nothing to install | CLI tool you install and run locally |
| Editing messages / tool results | Yes — in place, with validation | No — read-only export |
| Diagnosing a session that won't resume | Yes — Doctor lists current structural findings with click-to-jump; some repairs remain manual | No |
| Search & filter the timeline | Yes, with highlights | Reads as a replay page |
| Sharing a session with someone | You'd export the file itself | Its best feature — self-contained HTML |
| Privacy | File never leaves the browser | Runs locally too |
| Price | Free; payments are not live | Free, open source |
Which one do you need?
Your session won't resume — SessionPatch can locate malformed lines, missing parents, and unmatched tool events. Supported edits and linked-pair removal are local; some parent-chain repairs remain manual.
You want to change what a session says (remove a wrong turn, redact secrets, fix a tool output) — SessionPatch's editor.
You want to show someone what the agent did — claude-replay's HTML export is the nicer artifact. It's a good tool; this page isn't a knock on it.
You want session analytics (token counts, cost across sessions) — neither; that's agentsview territory.
Using both
Inspect or edit a copy in SessionPatch first, validate it with Claude Code, then replay the verified file. They compose rather than compete.