On August 26, Anthropic pulled two platform features out of beta that nobody’s cheering loudly about – but that change a lot for teams running Claude inside a company.
First up is the Compliance API. Its session endpoints are now official for Cowork and Claude Code, no longer beta. In plain terms: companies can pull and archive the transcripts of those sessions – the basis for audits, governance, and legally sound record-keeping. Newly added in beta are sessions from Claude Science and from Claude for Microsoft 365, meaning work done in Excel, PowerPoint, Word, and Outlook. It all runs through the existing Compliance Access Key with the right read scope.
Second is the Admin API, and this one is more for developers. Until now you had to poke at it via curl. Now it’s available in the ant CLI and in eight SDKs: Python, TypeScript, C#, Go, Java, PHP, and Ruby – under client.beta.organization. That lets you manage organization data, members, invites, workspaces, API keys, rate limits, service accounts, and even customer-managed encryption keys programmatically. A few things stay curl-only for now, like usage and cost reports and Claude Enterprise user management.
Why does this matter if you’re not an IT admin? Because it shows the direction of travel. Cowork and Claude Code stopped being playgrounds for individuals a while ago. When Anthropic starts serving audit-ready session transcripts and wrapping administration neatly into SDKs, it’s getting ready for serious corporate use – with everything legal and security teams demand there.
To me this is a good example of a second layer running under the visible feature updates. The models get the headlines. But whether a tool actually lands in a company often comes down to exactly these unglamorous questions: can I trace what happened, and can I manage it automatically? Since August 26, the answer is a little more often: yes.
Source: Claude Platform Release Notes