MCP Server

memories.sh includes a built-in MCP (Model Context Protocol) server that allows AI tools to read and write memories directly over stdio, without generating intermediate files.

Starting the Server

memories serve

The server communicates over stdio using the MCP protocol. It logs diagnostic information to stderr.

Why Use the MCP Server?

While memories generate creates static files, the MCP server provides live access to your memories:

• AI agents can search for relevant context based on the current task
• Agents can add new memories as they learn about your project
• No need to regenerate files when memories change
• Works alongside generated files — use both approaches together

Connecting to AI Tools

Add the MCP server to your AI tool's configuration. The exact setup varies by tool — see the Integrations guide for specific instructions.

The general pattern is to add memories.sh as an MCP server with the command memories serve (or npx @memories.sh/cli serve if not installed globally):

{
  "mcpServers": {
    "memories": {
      "command": "memories",
      "args": ["serve"]
    }
  }
}

Available Tools and Resources

The MCP server exposes:

• 7 tools — for reading, writing, searching, and managing memories
• 3 resources — for accessing rules, recent memories, and project-specific data

See the Tools Reference and Resources Reference for complete documentation.