Skip to main content

Connect Connie to Claude, ChatGPT and other AI tools

Use Connie from your AI assistant: set up Connie in Claude, ChatGPT or Claude Code, choose what agents may do, and see what they can't.

Connect Connie to the AI tool you already use, and ask it to work in Connie for you: find a document, draw up a contract from one of your templates, send it for signature or remind the people who haven't signed yet. Connie works with Claude, ChatGPT, Claude Code and any other AI tool that supports MCP (the Model Context Protocol).

Availability: Connecting AI tools is included in the Studio and Professional plans.

What you can ask

Once Connie is connected, you talk to your AI tool the way you normally would. A few examples:

  • "Which of our NDAs are still waiting for a signature?"

  • "Draw up a freelancer agreement from our template for Jane Doe, [email protected], starting 1 November, and send it for signature."

  • "Remind everyone who hasn't signed the Summer Festival release."

  • "Set up a SignPage for our photo consent template."

  • "What does the Acme contract say about termination?"

  • "Create a new project for Northwind."

Before the AI tool does anything that sends an email, publishes something or can't be undone, it shows you what it is about to do and asks you to confirm.

Before you start: turn on agent access

Agent access is off until somebody in your workspace turns it on. You need the Manage settings permission to change it (see Permissions explained: Manage settings).

  1. Go to Settings → AI & Agents.

  2. Turn on MCP Access.

  3. Decide what agents may do with your contracts. Both switches are off until you turn them on:

    • Read documents allowed: agents can read your contracts, including their text, where each one stands, and the names and contact details of the people signing them.

    • Send documents for signature allowed: agents can issue documents in your workspace's name, send them for signature and remind signers. With only this switch on, an agent can prepare and send new documents but can't open ones that have already been sent.

On the same page, the Setup Guide shows the address you'll need in the next step. It is:

https://mcp.getconnie.com

Connect Claude

On a Pro or Max plan:

  1. In Claude, go to Customize → Connectors.

  2. Click +, then Add custom connector.

  3. Enter https://mcp.getconnie.com as the server URL and click Add. Leave the advanced settings empty.

  4. Click Connect. Claude opens Connie, where you log in and approve the connection (see Approving the connection below).

On a Team or Enterprise plan, an owner of your Claude organization adds Connie first, under Organization settings → Connectors → Add → Custom → Web, with the same URL. Each member then finds Connie under Customize → Connectors and clicks Connect.

Connect ChatGPT

ChatGPT connects to Connie through its Developer mode, which is available on the web for Plus, Pro, Business, Enterprise and Education accounts. On a Business or Enterprise workspace, your ChatGPT admin may need to allow Developer mode first.

  1. In ChatGPT, go to Settings → Security and login and turn on Developer mode.

  2. Go to chatgpt.com/plugins and click + to create an app.

  3. Name it Connie, enter https://mcp.getconnie.com as the MCP server URL, and choose OAuth as the authentication.

  4. Log in to Connie and approve the connection when ChatGPT asks.

  5. In a new chat, pick Connie from the Developer mode tools in the message box.

Good to know: asking ChatGPT or Claude in a chat to "install" Connie from its address doesn't work. Neither can add a connector from inside a conversation, so it has to be added in settings as described above.

Connect Claude Code

  1. In a terminal, run:
    ​claude mcp add --transport http connie https://mcp.getconnie.com

  2. Start Claude Code, run /mcp, select connie and choose to authenticate. Your browser opens Connie, where you log in and approve the connection.

Other AI tools that support remote MCP servers with OAuth connect the same way: give them https://mcp.getconnie.com and approve the connection in Connie.

Approving the connection

When you connect an AI tool, Connie asks you to log in and shows which tool is asking for access. If you belong to more than one workspace, you choose which one to connect. The screen tells you what the agent will be able to do there, including whether it can read and send contracts, based on the switches above.

Each connection reaches one workspace. To use an AI tool with another workspace, or with a sub-account, connect it again and choose that workspace.

What an agent can and can't do

  • It acts as you. An agent can only do what you can do in Connie. If you can't see a project or edit a template, neither can the agent. If your permissions change, the agent's change with them straight away.

  • By default, it can work with templates, projects, SignPages and SignTags, and with anything else your permissions allow. Reading and sending contracts each need their own switch, as described above.

  • It asks before it acts. Sending documents, reminding signers, publishing, deleting and other actions that email somebody or can't be undone show you a summary and wait for your confirmation.

  • It can never sign anything. Signing always happens by the signer, in Connie.

  • Some things stay in Connie. An agent can't change Single Sign-On, API keys, billing or payment details, transfer account ownership, open signing links, or change the AI & Agents settings themselves.

Privacy

An AI tool you connect is not one of Connie's subprocessors. It is a service you choose, under your own agreement with its provider. If you let it read contracts, the personal data in them is shared with that provider on your instructions. You are responsible for having a lawful basis, a data processing agreement where one is required (including for transfers outside the EEA), and for telling the people concerned in your privacy notice. Check whether the provider may use your data to train its models.

Connie keeps a record of everything a connected agent does, including who connected it and when your settings changed.

See and disconnect agents

Under Settings → AI & Agents → Connected Agents you can see every AI tool connected to your workspace, who connected it, and disconnect it. You can always disconnect your own; disconnecting somebody else's needs the Manage team permission.

  • Turning off MCP Access pauses every connected agent. Their connections are kept, so they work again when you turn it back on.

  • Turning off a contract switch takes effect on the agent's next request, without reconnecting.

  • When a member is deactivated, the agents they connected stop working too.

Limits

  • Each connection can make up to 120 requests a minute, and a workspace up to 600 a minute across all its connections. That is far more than a conversation needs.

  • Documents an agent imports as templates can be up to 1 MB, and count towards the same daily import allowance as imports in the template editor.

  • An agent reads long documents a section at a time.

Troubleshooting

  • "This workspace has not turned on access for AI agents": somebody with Manage settings needs to turn on MCP Access under Settings → AI & Agents.

  • "This Connie workspace is on a plan that does not include AI agent access": connecting AI tools is part of the Studio and Professional plans.

  • The agent says it can't read or send a contract: check the Read documents and Send documents switches under Settings → AI & Agents.

  • The workspace you want isn't offered when you approve: MCP Access may be off there, or your membership may be deactivated.

Still stuck? Write to us at [email protected] or use the chat in Connie.

Related Articles
Did this answer your question?