Skip to content
FAQ

The three questions everyone asks first

  • Privacy

    Does any of my data leave my machine?

    Nothing leaves silently. The daemon listens on 127.0.0.1 only, and indexing, search and context re-injection are fully local. Every outbound call is user-initiated or opt-in and off by default, apart from the license activation and revocation check that come with a Pro license. The README network-activity table lists every one with its exact payload, and /telemetry covers the opt-in install ping in full.

  • Pricing

    Is Pro a subscription?

    No. One-time. $49.69, lifetime, never a subscription. Pay once, own it for life.

  • Future-proofing

    Will Anthropic just ship this natively?

    Maybe. The moat is recall context, threads, MCP, redaction, cross-AI support, and the dead-man clause. Anthropic is a cloud company. They cannot ship "even if we disappear".

Everything else

Tagged by topic so you can scan to the answer you need.

  • Privacy

    Can I use this if I am offline?

    Yes. Works on a plane, behind firewalls, on air-gapped machines. The browser talks to http://127.0.0.1 — your own laptop.

  • Privacy

    What happens if I delete a JSONL from ~/.claude/projects/?

    The SQLite copy persists. Recall only adds and updates — it never deletes session rows because the source file disappeared. Your indexed memory outlives the source files.

  • Privacy

    How big does the database get?

    About 800 MB for 1,100+ sessions and 150,000 messages in our reference environment. Storage is cheap; this is rarely a problem. Compression is on the roadmap if it ever becomes one.

  • Compatibility

    Codex, Cursor, Cline, Windsurf?

    Codex, yes. Codex sessions are indexed beside Claude Code, carry a Codex badge, import the Codex thread title, and filter by a Source control; tab correlation runs on macOS and Linux. Recall reads only ~/.codex/sessions/**/rollout-*.jsonl and ~/.codex/session_index.jsonl, refuses symlinks, never opens auth.json, config.toml or any Codex database, and makes no network requests of its own. Cursor, Cline and Windsurf: not yet, those tools write neither ~/.claude/projects/ nor a Codex rollout file.

  • Compatibility

    Multiple machines on one index?

    Not yet. Each machine has its own ~/.recall/db.sqlite. Team sync is on the roadmap — opt-in, end-to-end encrypted, still local-first.

  • Licensing

    What is the license model?

    Free tier: index, list, show, basic browsing. Pro tier ($49.69 one-time, lifetime): full-text and semantic search, context re-injection, MCP, session lookup, full web UI. The license validates at activation and, while online, re-checks once a day whether the key was revoked; offline it works forever, since a key is only disabled when our server explicitly says so. Dead-man clause: if our servers disappear, existing keys stay valid.

  • Licensing

    Refunds?

    A 30-day satisfaction guarantee from license activation. Email and we refund. The clock starts when you run recall activate, not when you click Buy.

FAQ

The questions that come up most:

#How big does the database get?

~800MB for 1,100 sessions and 150,000 messages in our reference setup. Storage is cheap.

#Multiple machines, same index?

Not yet. Each machine has its own ~/.recall/db.sqlite. Team sync is on the roadmap (opt-in, e2e encrypted, still local-first).

#What if I delete a session JSONL?

Recall keeps its SQLite copy. We add and update; we never delete on source-disappearance. Indexed memory outlives the source files.

#Codex?

Yes, and it is on by default with nothing to configure. Codex sessions are indexed alongside Claude Code, with a Codex badge, the imported Codex thread title (a name you set yourself still wins), tokens priced from OpenAI list rates, tab correlation on macOS and Linux, and a Source filter (All / Claude Code / Codex) in the UI, over MCP, and as --source on recall list, recall search and recall locate. Threads and search cover both agents at once.

Nothing runs if you don't use Codex. To turn the source off when you do, run recall codex off, which writes sources.codex.enabled for you; recall codex status reports what is in force, and recall doctor prints rollout and ingest counts.

Recall reads only ~/.codex/sessions/**/rollout-*.jsonl and ~/.codex/session_index.jsonl (thread titles), opens them without following symlinks, and refuses a symlink rather than reading it. It never opens auth.json, config.toml, hooks.json, history.jsonl, the ipc/ directory or any Codex SQLite database, never writes under ~/.codex, and adds zero new network egress.

#Cursor / Cline / Windsurf?

Not yet. Those tools write neither ~/.claude/projects/ nor a Codex rollout file. The source adapter added for Codex is the seam they would plug into.

#What's the license model?

Free for the basics. Pro is one-time, $49.69, lifetime updates, dead-man clause. The license validates at activation and, while online, re-checks once a day whether the key was revoked; offline it works forever, since a key is only disabled when our server explicitly says so.

#Why are my session titles showing the first sentence instead of my tab name?

The tab-name pipeline broke between the editor extension and the daemon. Run recall doctor and read the Pipeline health section, it tells you which layer is at fault. Most common cause: the daemon updated to require authentication, the Marketplace extension hasn't auto-updated yet. Reinstall the extension, run "Developer: Restart Extension Host", and the Auth rejections counter stops climbing the moment the new bundle takes over.

Last updated