Continue: MCP setup

Connect Continue to mountable.io to list filesystems, inspect usage and manage mount sessions through MCP.

The mountable.io MCP server runs locally over stdio. Install Node.js with npx available to the editor, then create an API key with access only to the filesystems you need.

Add the server

Create .continue/mcpServers/mountable.json in your workspace. Continue accepts the same JSON format as Cursor and Cline in this directory. Keep this file out of Git when you add a real key.

Put this in the file:

{
  "mcpServers": {
    "mountable": {
      "command": "npx",
      "args": ["-y", "mountable-cli@0.8.0", "mcp"],
      "env": { "MOUNTABLE_API_KEY": "REPLACE_WITH_YOUR_KEY" }
    }
  }
}

Replace REPLACE_WITH_YOUR_KEY in your private local configuration. Never paste the key into a chat, commit it, or pass it as a command argument. If npx is not on the editor’s PATH, use its absolute path.

Check the connection

Use Agent mode in the Continue extension to access MCP tools. Select the server in the MCP menu to inspect its tools. Continue CLI users can also inspect connected servers with /mcp. For a CLI config file, add the server under mcpServers in your existing YAML:

mcpServers:
  - name: mountable
    command: npx
    args: [-y, mountable-cli@0.8.0, mcp]
    env:
      MOUNTABLE_API_KEY: REPLACE_WITH_YOUR_KEY

Keep the rest of your existing configuration, including its model settings.

Ask the agent to call list_filesystems. The result contains the filesystems your key can access; an empty list is valid. Review the call before approving it. An authentication error means you should check the key and its grants, then restart the server.

The MCP tools manage filesystems, usage, mount sessions and webhooks. To read or write file contents, use a mount or the Python/TypeScript SDKs. Mounting remains a separate local command; wait for its mounted event before using the directory.

See the ten MCP tools, permissions, billing and Continue’s configuration guide.