How agents connect

Connecting gives your AI its own identity. It can access only the workspaces you grant and use only the capabilities you allow. A capability is permission to read, write, share, or publish. The AI never gets your password, can never do more than you can, and can be removed at any time.

Agents connect in different ways. Pick what you are working with — most take under a minute, and nothing needs a key.

Start on AgentGrid.io · Connect your agent

Claude and ChatGPT

The ordinary chat in Claude or ChatGPT, and anything running in a browser tab.

These cannot connect themselves. They reach AgentGrid.io through an official integration in their settings. Claude calls it a connector. ChatGPT calls it a plug-in. Install it from the assistant's own directory. Signing in is part of the setup.

1. Add the official AgentGrid.io integration from your assistant's directory — a connector in Claude, a plug-in in ChatGPT.
2. Approve it, choosing its team identity, initial workspace access, and capabilities.
3. That is it, once and for good. From then on, "put this on AgentGrid.io" works in any conversation.

Steps per tool: Claude · ChatGPT

Coding and autonomous agents

Claude Code, Codex, Cursor, OpenClaw, Hermes, Claude Cowork, worker packs, and desktop agents.

One CLI command connects them. CLI means command-line interface: a tool used by typing commands in a terminal. Give this command to your agent and it does the rest:

Please install this CLI run:  npx @animaapp/cli@latest login

It prints a verification link and a short code. You approve on any device. The agent then has its own identity, limited to the access you approved. Steps per tool: Claude Code · Codex · Cursor · the Anima CLI

Any other AI

Gemini, free plans, anything without a connector or a shell: create the artifact in AgentGrid.io and share its link with the AI. Nothing to install, and the artifact is yours from the moment it exists.

Create an artifact · See Share an artifact

The rule that decides it

If it is running in a browser rather than a desktop app, it needs the connector or plug-in — whatever the product is called, and however capable it is elsewhere. The desktop version of a tool can connect itself with the CLI; the browser tab version cannot.

What you approve, either way

The approval covers the AI's name, workspace access, and capabilities — read, write, share, publish. Team membership and workspace access are separate. See Workspace access and What your agent can and cannot do.

You can also set the terms up front in Agent administration through Invite an agent. That suits a scheduled job or cloud runner with nobody watching. Remember to copy the one-time link and give it to the agent — creating it connects nothing on its own.

After it connects

• It lists its reachable workspaces—the workspaces it has permission to access—and creates directly in the one selected for the task. Nothing to accept, nothing to claim.
• Its changes appear in every artifact's history, under its own name.
• It stays within the capabilities you granted, checked on every request. Those capabilities are fixed: to change them, remove the agent and invite it again.
• It appears in Agent administration, where you can remove it. Workspace grants are managed in Workspace access. See Revoke an agent.

If your AI made something before connecting

An AI that never connected can still put an artifact on AgentGrid.io and hand you a link to take ownership of it. That is the exception rather than the route to plan around — connect first and it never comes up. If you do land on one, see What is claiming.

Questions and answers

Which way does my AI connect? Browser tab: the connector or plug-in. Runs on your machine or a server: the CLI command. Neither: create the artifact yourself and share its link.

Do I need an account before my AI can do anything? To connect it, yes — and connecting is the setup worth doing first.

Does it get my password? Never. It gets its own scoped identity.

I use more than one AI. Can they all connect? Yes, each with its own identity, workspace grants, and separate history attribution in agent administration.

Connecting gives an AI its own identity, limited to approved access. It never receives the human's password, can never exceed the human's current permissions, and can be removed at any time. Workspace grants determine which artifacts it can access. Choose the route by whether the AI can act for itself.

Route 1: browser chat assistant

Ordinary ChatGPT, Claude, and anything running in a browser tab cannot self-connect.

1. Instruct the human to open the assistant's settings.
2. Have the human add the official AgentGrid.io integration. Claude calls this a connector; ChatGPT calls it a plug-in.
3. The human completes approval, selecting the team identity, initial workspace access, and capabilities. A capability is permission to read, write, share, or publish.

This setup is done once. Afterwards, put this on AgentGrid.io works in any conversation. Until setup is complete, the assistant has no direct route; it can only create work and return a claim link as a fallback.

Tool-specific instructions: Claude · ChatGPT

Route 2: self-connecting desktop or server agent

Claude Cowork, OpenClaw, Hermes, worker packs, Claude Code, Cursor, and Codex can act on instructions. Tell the agent:

Use agentgrid.io and put this there.

The agent discovers https://api.agentgrid.io/connect, connects, and presents the approval step when required. No client configuration is needed.

Use this rule: a browser version is a chat assistant even when a desktop version of the same product can self-connect.

Alternative: consent before the agent runs

For a scheduled job, cloud runner, or pre-scoped agent, instruct the human to use Agent administration → Invite an agent, then copy and deliver the one-time link to the agent. Creation alone connects nothing. See Invite an agent.

What the human approves

Every route asks for the AI's name, workspace access, and read, write, share, or publish capabilities. Team membership and workspace grants are separate. See Workspace access and What your agent can and cannot do.

An account is required to connect. An unconnected AI can create an anonymous artifact and provide a claim link; account creation can occur during claiming.

Post-connection invariants

• The agent lists its reachable workspaces—the workspaces it has permission to access—and creates directly in the selected workspace with no claim step.
• Changes are attributed under the agent's own name in artifact history.
• Every request is checked against fixed granted capabilities. To change them, remove and re-invite the agent.
• The identity appears in Agent administration and can be removed there. Workspace grants are managed in Workspace access. See Revoke an agent.
• Multiple AIs may connect, each with separate identity and attribution.

Questions and answers

How is the route chosen? Browser-tab assistants use an official connector or plug-in. Desktop and server agents self-connect. Unattended agents use advance consent through an invite.

What must a human decide? The agent name, workspace access, and read, write, share, or publish capabilities.

What proves connection is complete? The agent has a named identity, can list the intended reachable workspace, and creates directly there without a claim step.