Getting Started with Gatekeeper

Gatekeeper is a security layer that sits between your AI agents and the tools they use — like your CRM, cloud storage, or code repositories. It lets your agents take real actions across these tools without ever receiving or storing your credentials, and gives you full control over exactly what they're allowed to do.

Every action an agent attempts is checked against your policies before it runs, and every action — allowed or blocked — is logged in a complete audit trail.

What you'll do to get set up

  1. Connect your first service. From your Gatekeeper dashboard, go to Services and choose the tool you want your agent to access (e.g. HubSpot, GitHub, Slack).
  2. Authorize the connection. You'll be asked to log in to that service and approve the specific permissions (scopes) Gatekeeper is requesting. Only approve what your agent actually needs.
  3. Set your first policy. Policies decide what happens when an agent tries to take an action — allow it automatically, block it, or require manual approval. You can start simple (e.g. "allow all read actions, require approval for anything that deletes data") and refine over time.
  4. Test it. Have your agent attempt an action and confirm it behaves the way you expect. You can review exactly what happened in your audit log.

What happens after you're connected

Once a service is connected, your agent can only do what your policies explicitly allow. There's no ongoing manual step required — Gatekeeper evaluates each action automatically in real time.

If you ever want to pause everything immediately, you can disconnect a service or use the kill switch to instantly cut off agent access, no cleanup required.

Next steps