Skip to main content
Native hooks only capture sessions that happen after you install them. If you used Claude Code, Gemini CLI, Codex CLI, Cursor, or GitHub Copilot CLI before running rclm-hooks-install, those earlier sessions are still sitting in each tool’s local transcript files — rclm-sync finds them and uploads them into your RCLM history.
Historical sync is a one-time (or repeatable) backfill, not a replacement for hooks. Install hooks first so new sessions are captured going forward, then run rclm-sync to bring in everything that happened before.

​
Run a sync

By default rclm-sync scans every supported provider, shows how many sessions it found, and asks for confirmation before uploading anything. Pass --yes (or -y) to skip the confirmation prompt.
rclm-sync requires an API key. If you haven’t run rclm-hooks-install yet, do that first — the sync command reads the same ~/.reclaimllm/config.json credentials.

​
Where sessions are discovered

​
CLI flags

​
Scope to specific providers

Omit provider flags to scan everything (the default), or combine flags to scan a subset:

​
Skip the confirmation prompt

​
Re-upload everything

By default, rclm-sync tracks which files it has already uploaded in a local index and skips them on subsequent runs. To ignore that index and re-upload every discovered session:

​
Reprocess failed uploads

If a session fails to upload (for example, a network error), it’s quarantined locally instead of being dropped. Reprocess quarantined records with:

​
The sync index

rclm-sync tracks which session files it has already uploaded in ~/.reclaimllm/synced_sessions.json. On every run, newly discovered files are checked against this index so re-running rclm-sync never creates duplicate sessions in your dashboard. Use --resync when you specifically want to bypass this and re-upload everything — for example, after a parsing bug fix that changed how older sessions are converted.
Sync runs safely alongside live hooks capture. If a session file is picked up by both a running hook and a later rclm-sync pass, the sync index prevents it from being uploaded twice.

​
Troubleshooting

Confirm the tool’s local transcript directory actually contains files, using the discovery path from the table above. If the directory is empty, that provider simply has no local history to sync — this is expected on a machine where the tool has never been used.
Check whether it was already marked as synced in a prior run:
If the file’s path or ID appears there but the session isn’t visible in your dashboard, run rclm-sync --resync to force re-upload.
Failed uploads are quarantined rather than lost. Reprocess them once the underlying issue (network, API key, rate limit) is resolved: