An agent audit trail that comes with the memory

When an AI agent remembers something, acts, or is stopped by a rule, you need to know which agent it was, what happened, and whether anyone changed the record afterwards. With agentics.you the audit trail isn't a separate product. It's part of the memory. Every memory write and every rule decision is on the record as a receipt, on every plan, including Free.

Start free. No card to start.

What a receipt records

Each receipt holds:

• Who: the workspace and the agent.
• What: the action, and the version of the record.
• Under which rule: the rule and its decision, which is allow, deny or needs you.
• A fingerprint: a SHA-256 digest of the content. The receipt holds the digest, not your text.
• Where it sits: prev_hash, the receipt written just before it, so none can be slipped in or taken out.
• Provenance: verified, self-reported or imported, so you can see how much to trust each entry.

Recalls leave a receipt too. So does a deletion: DELETE /v1/memories/{id} returns a receipt_id.

How receipts get written

• Over MCP: remember returns a receipt_hash and a verify_url. Agents can also record their own work with record_action, log_action, report_outcome and record_outcome, and each of those returns a receipt hash.
• Over the API: search results from POST /v1/memories/search carry a receipt_hash on each hit. See the agent memory API.

Check any entry without an account

Paste a receipt hash at /verify/. Four checks run on it:

1. Unchanged. The receipt still gives the fingerprint it had when it was written.
2. In sequence. It points to the receipt written just before it.
3. In a sealed batch. Receipts are sealed into a batch every 30 minutes under one shared fingerprint.
4. On a public record. That shared fingerprint is anchored to a public record.

From code, GET /api/proof/verify?hash=… returns receipt_recomputed, chain_ok, merkle_ok, anchor_ok and verified. It never returns the memory's content.

What a check doesn't prove: it doesn't say the action was a good idea or the memory is true. It says the record wasn't quietly changed.

Limits as part of the record

Rules decide what an agent does on its own and what waits for your yes. Anything that waits goes to Needs you, and your answer is on the record as well. Rules, approvals and Pause all are on every plan.

Which plan gives you what

We price on actions, never per agent. Full pricing.

Start with one agent

Connect any MCP client to https://agentics.you/mcp (how), call remember, and open the verify_url it returns. That's your first audit trail entry.

Start free