Legal

Privacy Policy

Last updated: July 2, 2026

Total Recall is built local-first. Your memory stays on your machine. This policy explains exactly what the software stores, what leaves your computer, and what never does.

What Total Recall stores, and where

Total Recall indexes your AI agent sessions, the documents and folders you choose to connect, saved facts, and recurring work patterns. All of it is stored in a local database on your computer (by default under ~/.claude/total-recall/). Nothing is uploaded to Total Recall servers. There are no Total Recall servers that hold your data.

You can delete everything at any time by uninstalling, or by deleting the Total Recall home folder. Your data, your disk, your call.

What never leaves your machine

• Your sessions, conversations, and their content
• Your connected documents, notes, and knowledge files
• Your saved memories, decisions, and patterns
• Search queries and recall results
• The embeddings and indexes built from any of the above

Search and recall run entirely on your computer, including the embedding model, which is downloaded once and runs locally.

Network connections the software makes

Total Recall makes a small number of outbound connections, none of which carry your content:

• One-time model download. On first run, the software downloads a ~128MB open-source embedding model (bge-small-en-v1.5) from our content mirror (Cloudflare R2), with HuggingFace as a fallback. This is a plain file download. As with any download, the server can see standard connection metadata such as your IP address.
• Update checks. The software periodically checks GitHub for new releases and, by default, downloads and applies verified updates. GitHub sees standard connection metadata. You can disable automatic updates (TOTAL_RECALL_AUTO_UPDATE=0).
• License validation. The software periodically validates its license against our license service. This request contains an anonymized machine identifier (a one-way SHA-256 derivation of a platform identifier, not reversible to your hardware serial), the license status and tier, a timestamp, and a random nonce. It contains no personal information and none of your content.

That is the complete list. The software contains no analytics, no telemetry, no crash reporting, and no advertising identifiers.

Your AI provider

Total Recall works alongside AI agents you already use (such as Claude Code or Codex CLI). When those agents process your work, including context that Total Recall surfaces to them, that processing happens under your existing agreement with your AI provider, using your own account. Total Recall does not add any new party to that relationship and does not send your content anywhere on its own.

Background enrichment (summaries, digests) uses your own AI subscription in the same way, within the budget you configure.

The website

If you join the waitlist or sign up for updates at our website, we store the email address you submit so we can contact you about Total Recall. We do not sell or share it. To have it removed, email the address listed on the site or open a GitHub issue. The website also uses Cloudflare Web Analytics, which collects aggregate, non-identifying pageview data (such as device type, browser, and country) and does not use cookies or track you across sites.

Children

Total Recall is not directed to children under 16 and we do not knowingly collect personal information from them.

Changes

If this policy changes, the updated version will ship with the software and appear on this page with a new "Last updated" date.

Contact

Questions or removal requests: [email protected].