Latitude supports extending its capabilities by integrating with external tools and services through servers implementing the Model Context Protocol (MCP). This allows you to connect Latitude prompts to systems like Slack, Notion, databases, or custom internal tools.

​
What is an MCP?

MCP (Model Context Protocol) is a standardized way for AI platforms like Latitude to interact with external tools or “skills”. An MCP server acts as a bridge:

  1. Latitude Request: When a Latitude prompt (often an Agent) needs to use an external tool, it sends a request to the configured MCP server.
  2. MCP Server Action: The MCP server receives the request, understands which tool is being called and with which arguments, and executes the corresponding action in the third-party system (e.g., sending a Slack message, querying a database, creating a Notion page).
  3. MCP Server Response: The MCP server sends the result of the action back to Latitude.
  4. Latitude Continues: Latitude receives the result and continues the prompt execution (e.g., the agent uses the information retrieved from the database).

​
Setting Up an MCP Integration

The exact setup process depends on the specific MCP server you want to integrate.

​
Option 1: Using Pre-built MCP Servers

Several open-source or third-party MCP servers might exist for common tools (check community resources or the specific tool’s integrations).

  1. Deploy the MCP Server: Follow the server’s instructions to deploy it (e.g., as a container, serverless function).
  2. Configure in Latitude: In your Latitude project settings (or potentially self-hosted configuration), add the MCP server’s URL and any required authentication details.
  3. Enable Tools in Prompts: In your Latitude prompts, enable the specific tools provided by that MCP server in the tools configuration block.

​
Option 2: Building a Custom MCP Server

If you need to connect to a custom internal tool or a service without a pre-built MCP server, you can build your own.

  1. Choose a Framework: Use any web server framework (e.g., Node.js/Express, Python/Flask, Go/Echo).
  2. Implement MCP Endpoints: Implement the required HTTP endpoints specified by the MCP protocol standard (details typically found in Latitude’s advanced/developer documentation or community resources).
    • Handle incoming tool call requests from Latitude.
    • Parse tool names and arguments.
    • Interact with your target third-party service/API.
    • Format and return the results according to the MCP specification.
  3. Deploy and Configure: Deploy your custom server and configure it in Latitude as described in Option 1.

​
Example Use Cases

  • Slack Integration: An agent uses an MCP tool to send notifications or summaries to a Slack channel.
  • Notion Integration: A prompt uses an MCP tool to create new pages or query databases within Notion.
  • Database Query: An agent retrieves customer information from your internal database via a custom MCP server tool.
  • Calendar Management: A prompt schedules meetings using an MCP tool connected to Google Calendar.

​
Self-Hosting MCP Integrations

If you are self-hosting Latitude, you will also need to deploy and manage any required MCP servers within your own infrastructure and configure Latitude to communicate with them.

​
Next Steps

  • Explore available Community Integrations
  • Learn how prompts use Tools
  • Refer to specific MCP server documentation for detailed setup instructions.