Claude Enterprise Inference Hooks (Beta)

A security check that inspects every prompt and tool response before Claude acts on it, for companies on Claude Enterprise.

Every prompt and every tool response routes through a connection to your own security server, which returns an allow or deny verdict before Claude generates anything
Covers Claude chat, Claude Code, and Claude Cowork, including tool calls made through connectors, skills, and plugins
Admins point it at a data-loss-prevention tool they already run (Netskope, Palo Alto Networks, Proofpoint, Zscaler, or one built in-house) and set their own timeouts and failure tolerance
Rollout controls include a shadow mode that logs but always allows, role-based exclusions, and percentage-based rollouts
Closes the gap left by Claude Code's client-side hooks, which only checked activity on the user's own machine
In beta now for Claude Enterprise customers, configured from the organization settings

Enforcing the data rules your company already has across everything a team does in Claude, not just what people paste into chat.

Your security team blocked a wider Claude rollout because they had no way to inspect what employees send it.

Security and compliance leads at companies on Claude Enterprise.

Join the newsletter to get ahead with AI.

I share strategies, new modules, free guides, and tools to get you better at using AI in your work.