Some news sounds like a filing cabinet at first. This is one of those – and still worth understanding if you want to see how Cowork and Claude Code make their way into big companies. Anthropic has extended its Compliance API to cover Cowork and Claude Code too. Beta, starting with Enterprise customers.
What this is about
The Compliance API isn’t new. Security, legal, and compliance teams use it to pull activity from their Claude deployments – for audits, for eDiscovery, for internal oversight. Until now that mostly meant the classic Claude chats. Cowork and Claude Code weren’t part of it.
Now they are. And through the same interface those teams already use. New endpoints list an organization’s local sessions, fetch the metadata for a single session, and return its transcript. No separate logging per product, no new integration to build. Your existing Compliance Access Key does the job.
What’s inside a session
Each session gives the team two kinds of data. First the content: prompts, responses, tool calls, plus whatever skills and artifacts produced – all as transcript text. Second the metadata: verified user ID and email, organization ID, session and message IDs, timestamps.
The interesting part is that this reaches sessions running locally on employees’ machines. Cowork on the desktop, Claude Code in the terminal – work that used to live only on someone’s own computer now shows up cleanly next to the chats in the compliance feed.
There are a few limits. Claude Code on the web isn’t included, and neither are sessions run through the Claude Platform or through Bedrock, Vertex AI, and Microsoft Foundry. If you already export OpenTelemetry data, keep it running – the two work side by side just fine.
My take
This isn’t the news that gets applause on stage. But it’s exactly the plumbing that decides whether a regulated company clears Cowork and Claude Code at all. Without auditable logs, legal says no, and then the best agent feature in the world doesn’t help you.
There’s another side to it, though. If your Cowork and Claude Code sessions are now centrally retrievable, your agent work has become auditable – prompt by prompt. For compliance that’s a win; for how it feels to type, maybe a small chill. Both are true at once. If you use Claude at work, it’s just worth knowing this visibility is possible.
Sources: