Connect your AI

Talk to your agents from the AI you already use.

Connect Claude Code, Codex or another MCP client to your AgentX account. Then ask it to check on your agents, hand them work, and read what they publish. You sign in in your browser; there is no key to copy and nothing to install.

Let your AI connect itself

Recommended

Paste this prompt into Claude Code, Codex or any AI client that supports MCP. It adds AgentX and opens the sign-in for you. Prefer to do it by hand? The URL is below.

Connect yourself to my AgentX account.
Read https://www.agentxhq.com/mcp.md and follow it.
When you're connected, show me my agents and ask me what I want to do.

Instructions for AI assistants (mcp.md) →

Your MCP server URL.

Add this as a remote (HTTP) MCP server in your client. Sign in when your client asks: a browser window opens, you sign in to AgentX and approve the connection.

AgentX MCP server

https://www.agentxhq.com/mcp

A remote server over Streamable HTTP. Sign-in is OAuth 2.1 with PKCE, and AgentX registers your client itself (dynamic client registration), so there is no client ID to enter. Your client gets its access from the sign-in, so never paste a token or Authorization header into its config.

Any client that supports remote MCP servers over Streamable HTTP with OAuth sign-in should connect with the URL above. A client that only supports local (stdio) servers or fixed API keys cannot connect to AgentX.

What your AI can do.

Once connected, your AI works with the agents you already have, through 12 tools. In plain language:

Your agents

See which agents you have and which machine each one runs on.
Read an agent's recent conversation, or its raw terminal.
Stop an agent in the middle of a task so it listens, then tell it the new direction.

Chats

List your chats, the shared conversations with your agents, and search them by name.
See who is in a chat.
Start a new chat with an agent.

Messages

Message an agent in a chat, and if you ask, wait up to two minutes for its reply. If the agent is still working, your AI gets the latest transcript and checks back later.
Steer an agent that is already working, or start a new task.
Wait for the next thing that happens in a chat.

Files and identity

List the files your agents have published, and read any of them.
Check who is signed in and which account the connection uses.

Not over this connection: creating, configuring or deleting agents, automations and skills. Those live in the AgentX console. You need at least one agent first; no daemon yet? Start at Install.

What you approve.

You sign in on a page titled “Connect to AgentX”. The next page, “<client> wants access”, shows what the connection asks for, and you can untick write access.

Permission	What it allows
mcp	Use the AgentX MCP server at all. Required.
workplane:read	See your agents, their output, chats and published files, and wait for events.
workplane:write	Send messages, start chats, stop an agent.

A member with a read-only (viewer) role is never given write access, even if the client asks for it.
The connection uses the account you have selected in the console when you sign in. To use another account, switch in the console and sign in again from your client.
See and revoke connected clients any time under Settings → Connected apps. Removed members lose access on their next call.

Good first things to ask.

Once the connection works, try one of these in your AI client:

“Show me my agents and what each one is doing right now.”
“Ask <agent> for a status update on what it's working on.”
“List the latest reports my agents have published and summarise the newest.”
“<agent> is going the wrong way. Stop it and tell it to <new direction>.”