What this is

AI Control Center is the governed layer between AI clients and the systems where work happens — a surface of DevPlane. It lets you see every agent and what it touches, govern what it can reach and pause sensitive actions for a human, and prove what was accessed, decided, approved, and done.

Control starts before the agent runs. Finding a tool is not permission to use it — discovery checks availability; execution is where policy is enforced.

Who it's for

HR and enterprise-executive owners of AI risk — CHRO, CPO, Head of People Analytics, Total Rewards leaders, and the enterprise CAIO / AI-risk owner — plus the platform teams that build governed agents.

In three verbs

• See — one inventory of every agent, what it touches, who owns it, whether it's working.
• Govern — execution-time policy + connector scoping + an approval queue for sensitive actions.
• Prove — an evidence ledger (+ legal hold) where every action resolves to a control record.

Documentation

• Concepts — gateway · tool · registry · session · policy · confirmation · evidence
• Getting Started — issue a key, connect a client, make a governed call
• Architecture — components, data flow, and the substrate
• Gateway — the single governed front door
• Control Plane — governed actions, approvals, audit, legal hold
• Trust & Security — credential isolation, evidence, compliance posture
• API Reference — the gateway REST + MCP contract

Honest scope

This documents the target design; some substrate (multi-region residency, full self-host packaging) is roadmap. AI Control Center governs and records action — it does not make the business decision for you.