Documentation
The Agentic Control Plane sits in the path of every tool call your AI agents make. Each one is identity-checked, decided against your rules, priced, and audited — across every client and framework you run.
There are two ways in: put ACP in front of an agent you run (a hook, a wrapper, or the proxy), or run agents through the API and let ACP run the whole loop under the same rules. Pick where to start below.
Core docs
How a call is decided, the rules, identity, and how it all composes. The concepts behind every install — learn them once.
API
Create and run agents, set the rules they run under, and read the audit trail — from code. Full REST reference + OpenAPI spec.
Integrations
Install ACP into your AI client — Claude Code, Cursor, Cline, Codex, and more. One command, working in under five minutes.
Frameworks
Control agents you build — CrewAI, LangGraph, the Anthropic & OpenAI SDKs, Pydantic AI, and more. One wrapper.
New to ACP? Start from your own stack on Ways to set up ACP — one section per situation (Claude Code, Codex, Hermes, LangGraph/CrewAI, direct API, existing gateway), each ending with the coverage you'll have. Or run the API quickstart, or just point the installer at your coding agent:
macOS · Linux · WSL
curl -sf https://agenticcontrolplane.com/install.sh | bashWindows · PowerShell
irm https://agenticcontrolplane.com/install.ps1 | iex