Skip to main content
Add Industrial MCP when you build apps or agents in an AI tool that supports the Model Context Protocol (MCP). The tool can then use the same query, time series, and document tools that Atlas AI agents use and query the industrial knowledge graph in Cognite Data Fusion (CDF) under your permissions. Connect once from the Data fusion or Atlas AI workspace instead of building a custom integration for each agent.
This is not the documentation MCP server. To connect an IDE to Cognite documentation, see Connecting your IDE to Cognite documentation.

​
Before you start

  • A CDF user account for the project you want to query. The Industrial MCP server uses the permissions of the signed-in user. See Sign in to CDF.
  • An MCP-compatible AI tool, for example, Cursor, Claude, or GitHub Copilot.

​
Add Industrial MCP

By default, the Industrial MCP server uses the data scope from your selected location.
1

Open the Data fusion or Atlas AI workspace

Sign in to CDF and open one of the workspaces.
2

Open Industrial MCP

In the lower left of the sidebar, select Industrial MCP.
3

Copy the configuration for your tool

Select your AI tool, then copy its JSON configuration. The configuration uses the cluster endpoint, https://{cluster}.cognitedata.com/mcp. You select the CDF project during sign-in.
4

Add the server in your AI tool

Follow the instructions in the tool’s documentation to add the server.
5

Select a CDF project

When the tool starts the sign-in flow, select one CDF project on the cluster. The connection stays bound to that project. To query a different project, disconnect the server and sign in again.
6

Confirm the server is listed

Open the list of MCP servers in your tool.
The Industrial MCP server is listed and enabled. You can ask the tool a question about industrial data in your CDF project.
Last modified on October 28, 2026