home / mcp / master brain mcp server

Master Brain MCP Server

Central MCP server linking AI tools with Master Brain and Auto-Memory for persistent learning.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "activi-ai-master-brain-mcp-hub": {
      "command": "node",
      "args": [
        "/path/to/master-brain-mcp/dist/index.js"
      ],
      "env": {
        "BRAIN_API_URL": "http://178.156.178.70:3001",
        "ANTHROPIC_API_KEY": "sk-ant-xxxxx",
        "AUTO_MEMORY_ENABLED": "true",
        "AUTO_MEMORY_MIN_CONFIDENCE": "0.7"
      }
    }
  }
}

You are setting up a centralized MCP hub that connects multiple AI tools, stores important interactions automatically, and provides a robust runtime for continuous operation. This hub acts as the core brain, enabling cross-tool memory, efficient knowledge sharing, and automated learning across Claude Code, Codex, Cursor, and related components.

How to use

You deploy a local or remote MCP client to connect to the Master Brain MCP Hub. The hub exposes a local stdio-based interface when run as a server, or can be accessed remotely via an SSH-based workflow. Your client will send requests to the hub, which analyzes interactions with Haiku analysis, decides if the data should be stored, and updates the Master Brain with embeddings and contextual knowledge. You can query the brain, save knowledge manually, and inspect health and memory status to ensure smooth operation.

How to install

Prerequisites you need before installation: a machine with Node.js installed (recommended LTS version) and access to a shell. You will clone the project, install dependencies, configure environment variables, and build the project before running.

git clone https://github.com/Activi-AI/Master-Brain-MCP-Hub.git
cd Master-Brain-MCP-Hub
npm install
cp .env.example .env
# .env anpassen
npm run build

Additional setup notes

To connect from Claude Code, you use an MCP server entry that runs the hub locally or remotely. Example configurations below show how to point Claude Code to a local node process that serves the hub, including environment values for API URLs and keys.

Available tools

brain_search

Search the brain data store using semantic, keyword, or hybrid methods to locate relevant knowledge.

brain_save

Manually save knowledge or insights into the brain for future reference.

brain_stats

Display statistics about the brain data, such as entry counts and activity metrics.

brain_recent

Show the most recent brain entries to monitor recent activity.

brain_health

Check the health of the brain connection and the integrity of the brain data store.

auto_memory_status

Report the current status of the Auto-Memory middleware.

auto_memory_config

Configure Auto-Memory settings and behavior for automatic capture.