A Model Context Protocol implementation for the MemOS API service that provides conversation memory management and message processing capabilities for AI applications.

Installation

NPM Global

npm install -g @memtensor/memos-api-mcp

PNPM Global

pnpm add -g @memtensor/memos-api-mcp

NPX Direct

npx @memtensor/memos-api-mcp

Configuration

MCP Client

{
  "mcpServers": {
    "memos-api-mcp": {
      "command": "npx",
      "args": ["-y", "@memtensor/memos-api-mcp"],
      "env": {
        "MEMOS_API_KEY": "your-api-key",
        "MEMOS_USER_ID": "your-user-id",
        "MEMOS_CHANNEL": "the-site-where-you-are-seeing-this-document"
      }
    }
  }
}

Available Tools

Tool	Description
add_message	Adds a new message to a conversation with role and content information
search_memory	Searches memories in a conversation using a text query with customizable result limits
get_message	Retrieves messages from a specific conversation

Features

• MCP-compatible API interface
• Command-line tool for convenient interaction
• Built with TypeScript for type safety
• Express.js server implementation
• Schema validation with Zod

Environment Variables

Required

• MEMOS_API_KEY - Your Memos API key for authentication
• MEMOS_USER_ID - A stable identifier for the user that is deterministic and does not contain personal data; must remain the same across all devices/sessions
• MEMOS_CHANNEL - The site where you are seeing this document (e.g., MODELSCOPE, MCPSO, GITHUB, etc.)

Notes

Requires Node.js >= 18. Current version 1.0.0-beta.2. MEMOS_USER_ID should be generated using SHA-256(lowercase(trim(email))) or SSO subject/employee ID - never use random values or device identifiers.