Start here
Choose a Contro1 connection path: five agent setup guides for production agents, Claude Code, Codex, Cursor and personal agents, plus web sign-in for Grok, ChatGPT and Claude.ai. Follow exact steps and verify access.
Updated Oct 29, 2026
Pick the page for the kind of agent you have. Each one ends with the agent able to ask people for approval and run the application Actions you gave it, with every call checked and recorded.
Choose one of five agent setup guides, or use the app sign-in path for Grok, ChatGPT and Claude.ai. Each guide shows what to do and how to verify the connection.
| Kind of agent | Examples | Acts as | Steps |
|---|---|---|---|
| 1. Agents in production | LangGraph, Mastra, OpenAI Agents SDK, CrewAI, Pydantic AI, Strands, n8n | The agent itself | https://contro1.com/docs/connect-an-agent/production |
| 2. Claude Code | Claude Code | The Claude Code agent itself | https://contro1.com/docs/connect-an-agent/claude-code |
| 3. Codex | Codex | The Codex agent itself | https://contro1.com/docs/connect-an-agent/codex |
| 4. Cursor | Cursor | The Cursor agent itself | https://contro1.com/docs/connect-an-agent/cursor |
| 5. Personal agents | OpenClaw, NanoClaw | Each agent itself | https://contro1.com/docs/connect-an-agent/personal-agents |
Agents in production · Claude Code · Codex · Cursor · Personal agents
Your own code, running autonomously in production: a backend, a worker or a scheduled job, on Google Cloud, AWS, Railway, Render, Fly.io, Vercel or your own server. For: LangGraph, Mastra, OpenAI Agents SDK, CrewAI, Pydantic AI, Strands, n8n. Acts as: The agent itself. Credential: An Agent key in the platform's environment variables.
An agent in your own code gets its own identity and an Agent Credential, a key bound to that one agent. You give it Actions in Access, and it calls them with the SDK or plain HTTP; every call is attributed to it and checked by Contro1.
In short: Create the agent: Agents → New agent → "An agent you are building for production"; Give it Actions in Access; Create its Agent Credential and store it as CONTRO1_API_KEY; Call an Action from your code.
Anthropic's coding agent, on your own computer. For: Claude Code. Acts as: The Claude Code agent itself. Credential: A key in the Contro1 service on the computer.
Claude Code connects as its own agent with one command, which also adds the Contro1 MCP server for company apps. Optional hooks add approvals for terminal commands and a trace of every tool call.
In short: Install the Contro1 CLI; Sign in; Connect Claude Code, approve at the link it prints, then restart Claude Code; Check every line; Optional: approvals for terminal commands (hook).
OpenAI's coding agent: the desktop app, the CLI in a terminal, or the IDE extension. For: Codex. Acts as: The Codex agent itself. Credential: A key in the Contro1 service on the computer.
Codex connects as its own agent with one command, which also adds the Contro1 MCP server for company apps; a Codex hook then makes chosen shell commands wait for a person in Contro1. Signing Codex in as you instead still works, as an alternative.
In short: Install the Contro1 CLI; Sign in; Connect Codex, approve at the link it prints, then restart Codex; Optional: approvals for terminal commands, in .codex/hooks.json.
The AI code editor's agent, on your own computer. For: Cursor. Acts as: The Cursor agent itself. Credential: A key in the Contro1 service on the computer.
Cursor connects as its own agent with one command, which adds the Contro1 MCP server for company apps. Signing Cursor in as you instead still works, as an alternative.
In short: Install the Contro1 CLI; Sign in; Connect Cursor, approve at the link it prints, then open a new Cursor window; Check every line.
Assistants that run on your own computer or server and answer you in chat apps. For: OpenClaw, NanoClaw. Acts as: Each agent itself. Credential: A key in the Contro1 service on that computer, which cannot be exported.
One command per computer connects every OpenClaw agent or NanoClaw agent group as its own identity, with you as its accountable owner. Commands and sensitive steps then wait for a named person in Contro1, and each agent can use the company apps it was given.
In short: Install the Contro1 CLI; Sign in; Connect, then approve the agents at the link it prints; OpenClaw: make commands stop for approval, then start the Contro1 bridge; NanoClaw: install the Contro1 channel, then let Contro1 approve; Check every line.
Whichever way you connect, the agent starts with nothing. It can ask people for approval and report what it did. To use an application such as Gmail, these three have to exist:
An identity with an accountable owner. contro1 connect creates it when its owner approves. For your own code, create it under Agents, New agent, or with contro1 init. A web or MCP sign-in (Claude.ai, or Cursor and Codex signed in as you) acts as the person who approved it.
An administrator turns the application on under Apps, for the organization, and chooses which of its Actions are allowed. That alone gives no agent a tool.
In Access, a person gives a specific agent specific Actions on a specific account. People can do this for their own agents, within what the organization allows. A personal account is used on behalf of the person it belongs to, who turns that on for the agent in Access.
Apps that add a remote MCP server with a web sign-in connect without the CLI. The app acts as you: it can do what you can do and nothing more, and every call is recorded as yours.
Your own code running autonomously in production (LangGraph, Mastra, OpenAI Agents, CrewAI, plain HTTP): Agents in production. Claude Code: Claude Code. Codex: Codex. Cursor: Cursor. OpenClaw or NanoClaw: Personal agents. Claude.ai, ChatGPT, Grok Bot or Grok Chat: add the MCP address on this page.
No. Connecting establishes who the agent is and who answers for it. Access to an application comes only from Actions somebody who may give access to that account grants it in Access.
Yes. contro1 connect cursor and contro1 connect codex make each one its own agent on this computer, with you as its owner, and add the Contro1 MCP server, the same way contro1 connect claude-code does. Signing in as you instead (contro1 mcp login --execute with contro1 mcp serve) still works, and then the app acts as you. Codex can also stop shell commands for approval with the contro1 hooks codex hook. The same command connects Antigravity, Gemini CLI, GitHub Copilot, Windsurf, OpenCode, Amp, Augment, Devin, Factory Droid, Kiro and Qwen Code: contro1 connect antigravity, contro1 connect gemini and so on. Claude Cowork adds Contro1 as a connector, under Customize.
OpenClaw, NanoClaw, Claude Code, Cursor and Codex: in the Contro1 service on the computer, as a key that cannot be exported; the agent never sees it. Cursor or Codex signed in as you instead: in the contro1 CLI's MCP sign-in. Your own code: an Agent Credential in your secret manager. Claude.ai, ChatGPT and Grok: an OAuth token held by the app.
Connect your own agent (LangGraph, Mastra, OpenAI Agents, CrewAI or plain HTTP) running on Railway, Render, Fly.io, Vercel, AWS or Google Cloud: give it its own identity and an Agent key per place it runs, store the key in the platform's environment variables, give it Actions in Access, and call them.
Connect Claude Code as its own agent with contro1 connect claude-code, which also adds the Contro1 MCP server, and add approvals for risky commands. Exact commands and what you should see.
Connect Codex as its own agent with contro1 connect codex, which adds the Contro1 MCP server for company apps, and stop chosen shell commands for a person with a Codex hook. Exact commands and what you should see.
Connect Cursor as its own agent with contro1 connect cursor, which adds the Contro1 MCP server to ~/.cursor/mcp.json for company apps. Exact commands and what you should see.
Connect OpenClaw and NanoClaw assistants to Contro1: one command per computer, owner approval, approvals for commands, and company apps over MCP.
Choose the applications your organization approves, decide who may do what, and connect one MCP. Every person and agent gets exactly their access, and every action is recorded on their name.
Install and use the contro1 CLI to register AI agents, create approval requests, push AI inventory, retrieve evidence and traces, and test workflows before using the SDK or API.
