AI agents
Deploy, review and investigate with your own Claude or OpenAI key.
Bring your own key
An admin adds an Anthropic (Claude) or OpenAI API key under Agents → AI provider. The key is verified, stored encrypted and never returned to the browser; usage is billed to your provider account.
Built-in agents
- Release manager — publishes, promotes and advances rollouts based on release health; rolls back when needed.
- Release reviewer — reviews updates (signing, runtime, failure rate) and recommends go / hold / roll back.
- Security sentinel — investigates security alerts, audit logs and API key use; proposes containment.
Guardrails
- An agent never has more rights than the member who asked it.
- Each tool is Allowed, Needs approval or Denied per agent; write tools need approval by default.
- Every tool call is recorded in the agent history and the audit log.