Skip to content
Contact Join Support
Login
  • There are no suggestions because the search field is empty.

Connecting Join to AI Applications

Join supports the Model Context Protocol (MCP), a way for AI systems such as ChatGPT, Claude and others to connect.

The Model Context Protocol is an open-source standard for connecting AI Applications to external systems. Join supports this protocol to enable AI applications such as ChatGPT, Claude, and custom-built agents/applications to leverage customers' data and Join functionality.

The MCP URL for Join is: https://api.join.build/mcp/join. You will need to provide this URL to your AI application.

The precise way to connect Join depends on the application. Some applications may require a particular subscription level or mode to be enabled. Please see instructions for connecting to these applications:

Frequently Asked Questions

How does security work with the MCP Server? Will other users get access to my data?

In short, no. The Join MCP server utilizes the same authorization and authentication logic as the web application. Upon connecting, users authenticate via OAuth. Following this, they have access to the same data in their AI application that they would have access to via the web. No more, no less.

Does using the MCP Server give my data to the AI companies?

It might. That depends on which system you connect the Join MCP Server to, and how it is configured. Many AI applications and LLM providers make it possible to configure which chats are persisted, and whether data provided to them can be used in future training. Ask your internal IT teams which tools are supported or forbidden.

Connecting Join data via the Join MCP Server is like pasting data into an AI chat interface. You should not do it with AI Applications that you do not trust.

How can I be sure that I will not accidentally destroy or corrupt data in Join?

Join's MCP server supplies several tools for querying and modifying data in Join. Before invoking one of the tools, most AI Applications will confirm the action, and will often label whether it is a read-only action or could potentially modify underlying data. This keeps the user in control of every action that the LLM takes on their behalf via the Join MCP Server.