Model Context Protocol (MCP)

Model Context Protocol (MCP) extends Mind's capabilities by connecting to specialised servers that provide additional tools and context. This guide helps you set up, configure, and use MCP servers with Mind across all surfaces.

Capability	IDE	CLI	Web
Local (stdio) MCP servers	✓	✓	✓
Remote (HTTP/SSE) MCP servers	✓	✓	—
MCP config file (JSON)	✓	✓	✓
Server-provided prompts and resources	✓	✓	✓

What is MCP?

MCP is a protocol that allows Mind to communicate with external servers to access specialised tools, prompts, and resources. For example, a database MCP server provides tools to inspect schemas and run queries directly within Mind.

With MCP, you can:

Access specialised knowledge bases and documentation
Integrate with external services and APIs
Extend Mind's capabilities with domain-specific tools
Use server-provided prompt templates and resource templates via the # mention system in chat
Respond to server elicitation requests when tools need additional input during execution
Create custom tools for your specific workflows

Setting up MCP

Prerequisites

Before using MCP, make sure you have:

The latest version of Mind installed
Any specific prerequisites for the MCP servers you want to use (listed in each server's documentation)

Adding an MCP server

Open the command palette (Cmd + Shift + P on Mac, Ctrl + Shift + P on Windows/Linux) and search for Mind: Open workspace MCP config (JSON) or Mind: Open user MCP config (JSON).
Add your server configuration:

{
  "mcpServers": {
    "fetch": {
      "command": "uvx",
      "args": ["mcp-server-fetch"]
    }
  }
}

3. Save the config file (Cmd+S). Servers reconnect automatically. Check the MCP servers tab in the Mind panel to confirm the server is connected.

Installing from an install link

Some server directories and websites offer one-click install links that use the mind:// URL scheme. Opening one of these links doesn't change your setup immediately: Mind shows a confirmation dialog before anything is written to your configuration. The dialog lists the command that will run and its arguments, along with the names of any environment variables or headers — their values stay hidden. Review the details and confirm to add the server, or cancel to leave your configuration unchanged.

Agent configuration

You can also define MCP servers directly in an agent's configuration file. The mcpServers field specifies which MCP servers the agent has access to:

{
        "name": "myagent",
        "description": "My special agent",
        "mcpServers": {
          "fetch": {
            "command": "fetch3.1",
            "args": []
          }
        },
        "includeMcpJson": false
      }

The includeMcpJson field determines whether to include MCP servers defined in workspace and user-level configuration files.

Troubleshooting

Issue	Solution
Connection failures	Verify prerequisites are installed correctly
Permission errors	Check that tokens and API keys are valid
Tool not responding	Review MCP logs for specific error messages
Configuration not loading	Validate JSON syntax and save the config file