Turn old sessions into review, learning & sync
Your AI sessions are full of decisions, mistakes and lessons. relai mines them — locally — for risks you missed and concepts to study.
Scan sessions for what you missed
Scans ask Jev first — TypeSafe's small decision model answers typed questions about each slice of the transcript and points at the exact message as evidence, for a fraction of a cent per session. Only the slices it is not confident about go to your configured LLM. Findings show up as a coloured dot in your session list — or set scans to run on a schedule and get a desktop alert the moment something high-severity lands. One click opens your AI tool in the project folder to fix it.
API key pasted into the transcript
A live secret (sk-…a91f) was shared inline while debugging. Anyone with the session file has it.
- Rotate the exposed key and purge it from the transcript.
- Move secrets to a .env file the AI never sees.
- Secrets & API keys in the transcript
- SQL / command injection
- Auth & authorization bypasses
- Sensitive data in logs
Each session is one cached LLM pass — re-scanned only when the file changes.
Runs on a schedule
Flip on auto-scan and relai checks your recent sessions every few hours — fully local. Anything high-severity fires a desktop alert and a saved report, so a leaked key or a regression surfaces without you looking.
Fix it in one click
“Fix this” opens your AI coding tool — Claude Code, Codex, Cursor and more — right in the project folder that session ran in, handed the report to act on. No copy-paste, no hunting for the directory.
Every session is a lesson
A learning report distills a session into the concepts you actually leaned on the AI for — then builds a 2–4 week plan to close the gap. Preview it, download the markdown, or push it straight to your Obsidian vault.
Your sessions, on every machine
Opt into sync and the sessions you capture on one machine show up on the rest — encrypted before upload. The sync service supplies the encryption key to your signed-in devices.