Anthropic Launches Compliance API for Claude Code to Address AI Agent Governance Gaps

Anthropic has released a new Compliance API for Claude Code, giving security teams their clearest view yet into the activity of AI agents running on local developer machines. The API exposes detailed visibility into file reads, shell command execution, and invocations of MCP tools, along with the credentials available to the agent at runtime.

The release comes as AI has moved from the browser tab to the local development environment, where Claude Code reads files, runs shell commands, and invokes MCP tools through the credentials available on a developer's machine. While the new API endpoints provide improved visibility, Anthropic acknowledges that activity logs alone cannot tell security teams whether an agent's access is legitimate.

The company recommends that security teams review the documentation for configuration guidance, implement least-privilege access for Claude Code agents, and monitor activity logs in conjunction with other governance controls. No software patch is required, but configuration and policy updates are advised to mitigate risks associated with improperly configured agent access.