Configure an MCP server
AI Assistant can interact with external tools and data sources via the Model Context Protocol (MCP). An MCP server enables AI Assistant to use these integrations to process the context in your project, resulting in smarter, more relevant responses.
You can connect to your own MCP server, use an existing implementation from the official MCP repository, or import a Claude MCP config.
In the chat, type
/and select the Add Command option.
Alternatively, go to .
Depending on your setup, do one of the following:
Click
Add.In the New MCP Server dialog, specify the parameters of your MCP server:
Name – specify a name for the server.
Command – specify the command used to launch the server.
Arguments – specify additional parameters passed to the command.
Environment Variables – specify optional variables that your server might need to run (API keys or configuration settings).
Working directory – specify the path to the folder where your MCP server files are located.
Alternatively, you can provide your configuration as JSON. To do this, click Command in the top-left corner of the dialog and select the As JSON option from the list. Then, add your JSON configuration and specify the Working directory.
Click OK. Your MCP server will appear in the list.
Simply click
Import from Claude. All configured MCP servers will be added to the list. 
In the Level column, select whether the MCP server setup must be available globally or only in the current project.
Click Apply to save changes.
The status of your connection to the MCP server will be shown in the Status column. Once the connection is successfully established, you can review the list of available commands by clicking the icon in the Status column.
Commands available on the MCP server can be invoked in AI chat using the / symbol:
