Integration · Claude Enterprise

Signed, independently checkable evidence over your Claude activity.

Claude Enterprise records what your teams and agents did. Seal turns that record into evidence someone outside your organisation can verify: each audit and activity event mirrored, signed and committed to a public transparency log, checkable offline against a published key. The Activity Feed is authoritative, but it is vendor-held and mutable; Seal gives you a signed copy that stands on its own.

In preview. The connector is built to Anthropic’s documented Compliance API Activity Feed and needs only an Activity-Feed-scoped key to run. Request access to enable it on your organisation.

How it works

You enable the Compliance API
An organisation admin turns on the Compliance API and creates an Admin API key scoped to read:compliance_activities, the Activity Feed only. Seal never asks for a key that can read chats, files or session transcripts.
Seal reads the Activity Feed
The connector polls GET /v1/compliance/activities with a cursor, backs off on rate limits, and resumes from its last committed position, so no event is missed or double-counted.
Each event becomes a signed record
Seal signs a record carrying the event metadata and a SHA-256 of the event, never the conversation content. The actor is reduced to its type; email, IP and user agent are dropped. Every record is committed to Seal’s public, append-only transparency log.
A reviewer checks it offline
The evidence pack verifies in the reviewer’s own browser against a published key, with no account and no call to Aqta or to Anthropic. They can pin the log head and later confirm nothing was removed.

Why it is more than log shipping

Seal already holds a signed record of what your agents were authorised to do (its ACTION-v1 format, live in production). The Activity Feed holds what Claude actually did. Reconciling the two surfaces the one gap neither shows alone: an authorised action with no matching activity, or an activity with no matching authorisation. That is the check a reviewer actually wants, and it is only possible when an outside party holds both halves.

What it does, and does not, establish

This notarises Claude’s Activity Feed. It records that Aqta received these exact event bytes at this time; it does not re-run the underlying action and does not attest that the feed is complete.
It covers audit and activity events only. It does not ingest conversation content, and it does not imply the Compliance API covers model inference content on the Claude Platform.
It applies to Claude Enterprise via the Compliance API. It does not cover Claude accessed through Amazon Bedrock or Google Vertex AI.

Attribution and data handling

Every mirrored event and every derived evidence section is labelled as sourced from Claude by Anthropic. The Activity-Feed key is stored in a secrets manager, read only by the Seal gateway service on EU infrastructure, never in a browser and never logged. The evidence carries hashes and metadata, not prompt or response text, so verification retains no sensitive content. The specification, the conformance vectors and both reference verifiers are public; you can check the format itself before you enable anything. See the setup documentation.

Seal sits alongside Anthropic, not against it: Anthropic records the activity, Seal makes that record independently checkable. Checked rather than believed.