Mem-Brain MCP Server

MCP (Model Context Protocol) server that exposes Mem-Brain API functionality as standardized tools for AI agents. Built with FastMCP for production-ready HTTP/SSE transport.

Underlying HTTP API (tools, async jobs, search patterns): see Membrain API: Guide for Agentic Applications in the parent repo.

Features

• Memory CRUD and semantic search
• Search modes: raw (graph-native), interpreted (LLM summary), both (summary + evidence)
• Graph operations (path + neighborhood)
• Stats and health endpoints
• HTTP/SSE transport for remote MCP clients

Authentication

Mem-Brain MCP uses MemBrain API keys.

• Preferred header in MCP client config: X-API-Key: mb_live_...
• Also supported: Authorization: Bearer mb_live_...

Installation

PyPI

pip install mem-brain-mcp
mem-brain-mcp

uvx

uvx mem-brain-mcp

Configuration

API_BASE_URL=https://mem-brain-api-cutover-v4-production.up.railway.app
MEMBRAIN_API_KEY=mb_live_xxx   # optional fallback for single-user local runs
MCP_SERVER_HOST=0.0.0.0
MCP_SERVER_PORT=8100
LOG_LEVEL=INFO

MCP Client Examples

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "mem-brain": {
      "url": "http://localhost:8100/mcp",
      "headers": {
        "X-API-Key": "mb_live_xxx"
      }
    }
  }
}

Claude Desktop

{
  "mcpServers": {
    "mem-brain": {
      "url": "http://localhost:8100/mcp",
      "headers": {
        "X-API-Key": "mb_live_xxx"
      }
    }
  }
}

Development

pytest

Build / Publish

uv build
uv tool run twine upload dist/*