Pigment offers Model Context Protocol (MCP) solutions that Pigment users can leverage to access Pigment features securely through AI systems such as ChatGPT, Claude and Mistral.
What is MCP?
MCP is an open standard that allows AI systems to connect to external tools and data sources in a controlled, secure, and structured way.
By exposing Pigment capabilities via MCP, you enable AI systems to query data and retrieve insights directly from Pigment—all while respecting user permissions and access rights.
🎓 Pigment Academy
See also the Pigment Feature Spotlight videos on the MCP Server:
Before you begin
The MCP Server is available when you activate Pigment AI. Check whether MCP Server is already available by following the Enable MCP server step-through below. If you cannot toggle on MCP (Model Context Protocol), it may not be activated in your Workspace.
Configure data access rights
When a user accesses Pigment through the MCP server, they must authenticate with their Pigment account. All their existing permissions and access rights apply:
They can only access Applications, Metrics, or data they already have access to in Pigment.
No additional access rights or permissions are granted through MCP.
⚠️ Important
Before enabling MCP, verify that access rights are properly configured, especially on Blocks containing sensitive data.
To completely prevent a Block from being accessible via MCP and Agents, disable its “AI data access” setting.
Enable MCP server
ℹ️ Note
You need the Role of Security Administrator, Workspace Administrator or Primary Owner to enable MCP.
Go to your Workspace and select Settings from the sidebar.
Select Integrations from the sidebar then the MCP tab in the main pane.
Toggle on MCP (Model Context Protocol).
A unique MCP endpoint URL is generated, which you will need to connect a third-party AI system to Pigment.
If you plan to connect multiple Workspaces, each must have its own MCP endpoint URL and connector. Test & Deploy environments also need one MCP endpoint URL and connector each.
⚠️ Important
Once MCP is enabled and a connector has been created with the MCP URL, all Workspace Members can connect to the Pigment MCP server.
Troubleshooting
If you encounter issues while connecting:
Authentication fails:
In your Workspace settings, verify that the MCP server is enabled and that your MCP client is allowed.
Verify that you are using valid Pigment credentials, the correct MCP URL for the target Workspace, and that you have access to that Workspace.
Data leaks: check that data access rights are configured properly.
Data not visible: confirm that AI Data Access is enabled for the Metrics you want to query. Only Metrics with the following data types can be read via MCP: Number, Integer, and Boolean.
Modeling tools not visible: in your Account settings, check that advanced mode is enabled.