Skip to main content
Coconut works with the tools you already use, from Claude and ChatGPT to any client that speaks MCP.

Claude

Claude has a listing in Anthropic’s connector directory, so this is the shortest path if you’re using Claude as your primary client.

Set up in Claude

1

Open Customize in Claude

In Claude’s settings, open Customize.
2

Select Connectors

Go to the Connectors tab.
3

Add a connector

Select Add, then Browse Connectors.
4

Find Coconut

Search for “Coconut”, or open the directory listing directly.

ChatGPT

ChatGPT reaches custom MCP servers through developer mode rather than its plugin directory, so this mirrors OpenAI’s own setup flow.

Set up in ChatGPT

1

Turn on developer mode

In ChatGPT, open Settings, then Security and login, and turn on Developer mode.
2

Add a connector

Go to chatgpt.com/plugins and select the plus button.
3

Name the connection

Enter a name and description for Coconut.
4

Paste the connector URL

Under Connection, paste the connector URL from Integrations in the app.
5

Create and sign in

Create the connection, then sign in when prompted.

Microsoft 365 Copilot

Copilot Studio’s MCP wizard handles the OAuth handshake on its own, since the connector advertises dynamic client registration.

Set up in Microsoft 365 Copilot

1

Open Tools

Open your agent in Copilot Studio, then go to Tools.
2

Add a new tool

Select Add a tool, then New tool, then Model Context Protocol.
3

Name the server

Name the server and paste the connector URL as the Server URL.
4

Choose OAuth

Choose OAuth 2.0 with dynamic discovery, then select Create.
5

Add to your agent

Create a connection, then select Add to agent.

Gemini Enterprise

How to connect Coconut to Gemini Enterprise.

Set up in Gemini Enterprise

1

Issue a client

In Connector Governance, issue a client with the Gemini Enterprise preset, and keep the values it shows.
2

Open Connected apps

In Gemini Enterprise, open Manage team, then Connected apps, then Add MCP Server.
3

Add the server URL

Paste https://api.coconut.md/mcp as the server URL, then choose OAuth 2.0.
4

Paste the client details

Paste the authorization URL, token URL, client ID, and client secret from Connector Governance.
5

Add the scopes and sign in

Enter the scopes context:read context:write, select Add, then sign in when Gemini prompts you.

Cursor

Set up in Cursor

1

Open custom MCP settings

Open Customize in the sidebar, then add a custom MCP server.
2

Add the connector URL

Use the connector URL from Integrations in the app as a streamable HTTP server.
3

Sign in

Sign in when Cursor prompts you to authorize.

Grok Bot

Grok Bot connects to Coconut with an agent key, and you set it up by asking your bot in chat.

Set up in Grok Bot

1

Create an agent key

In Coconut, open Manage agent keys and create a key. Your bot can only reach the spaces that key can access.
2

Ask your bot to connect

In your bot’s chat, say: “Connect the Coconut Context connector at https://api.coconut.md/mcp using my API key.”
3

Approve and enter the key

Confirm when the bot asks, then enter your key in the secure field it shows you.
4

Check the connection

Ask your bot which Coconut spaces it can see. If a space is missing or shows Forbidden, ask an admin to give your key access.
5

Give access to private spaces

To let your bot into a private space, add its agent key as a member at https://app.coconut.md/spaces/<space-path>/manage/members/add/agent, for example agent-olivia. Pick the agent key from the list and choose the role it needs.

Something else

Any other MCP-compatible client connects the same way, and a custom integration or an automation runtime can use the same connector URL and agent token directly. See Connect your agents for that setup and the full configuration reference.