Skip to content

Quick Start

Get MidOS running with your AI agent in 3 steps.

  • Python 3.10+ installed
  • An MCP-compatible client: Claude Desktop, Cursor, Cline, VS Code, Windsurf, Zed, or OpenCode
  1. Install the package

    Terminal window
    pip install midos-mcp
  2. Configure your MCP client

    Edit ~/AppData/Roaming/Claude/claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

    {
    "mcpServers": {
    "midos": {
    "command": "midos-mcp",
    "args": ["serve"]
    }
    }
    }
  3. Verify the connection

    Terminal window
    midos-mcp health

    You should see:

    MidOS Health Check
    ==================
    Status: HEALTHY
    Knowledge: 21346 chunks | 118 skills
    Vector DB: Connected (290000 vectors)

Once connected, try these in your agent:

Use the search_knowledge tool to find information about "React Server Components"
Use the list_skills tool with stack "python" to see Python-related skills
Use the hive_status tool to see knowledge base metrics
Use the agent_handshake tool with your model name and project goal to get a personalized tool catalog

The free tier works without authentication (9 tools, 50 queries/day). For full access:

Terminal window
# Set your API key as environment variable
export MIDOS_API_KEY="midos_sk_dev_your_key_here"

Or pass it in the MCP config:

{
"mcpServers": {
"midos": {
"command": "midos-mcp",
"args": ["serve"],
"env": {
"MIDOS_API_KEY": "midos_sk_dev_your_key_here"
}
}
}
}

See Access Tiers for details on what each tier unlocks.