February 27, 2026
AI giveth, AI taketh away
Show HN: Claude-File-Recovery, recover files from your ~/.claude sessions
AI “cleaned” your files? This tool digs them back up — and the comments are chaos
TLDR: A new tool reconstructs files from Claude’s own logs, letting devs recover work after AI mishaps. Commenters split between praising life‑saving recoveries, questioning why such logs exist, demanding a built‑in “rewind,” and warning that flaky exports elsewhere have already cost people days—making reliable recovery a must-have.
Dev panic meets digital necromancy: “Claude-File-Recovery” claims it can resurrect files that AI assistants created or edited, by replaying Claude’s hidden session logs. In plain English: Claude keeps a diary of every file it touched, and this tool rewinds that diary so you can get your stuff back, view changes, and export them fast.
But the community drama is the real show. Some users are cheering because they’ve already been saved. One dev says “AI ran a git clean on me” and then Claude basically recovered the mess by reading its own transcripts. Another insists Claude itself found the files without extra tools—just tell it to look in its other chats. Cue the debate: do we even need this tool, or should this be a built‑in “/rewind” button?
Others are side‑eyeing the whole thing. Why is Claude keeping that much data in a secret-ish folder? One commenter is furious at a rival app’s exports, calling them a “fraudulent scam” after losing days of work when half the session went missing. Meanwhile, the practical crowd suggests old‑school safety nets like macOS snapshots. The vibe: AI is both arsonist and firefighter, and this tool might be the extinguisher with receipts. Also, yes, devs are joking about a new command: /rewind my life.
Key Points
- •Claude-File-Recovery reconstructs files from Claude Code by replaying Write, Edit, and Read operations recorded in JSONL logs under ~/.claude/projects/.
- •It features an interactive TUI with fuzzy search, vim keybindings, colored diffs, batch extraction, and smart-case search.
- •The tool supports point-in-time recovery and can reconstruct files as they existed before a specified timestamp.
- •Performance optimizations include parallel parsing with orjson and a fast-reject byte check that skips about 77% of lines.
- •Installation is available via uv (recommended), pipx, or pip, requires Python 3.10+, and the project is MIT-licensed and open to contributions.