MuntuAI MCP

This repository contains the public documentation, manifest, and reference clients for the live MuntuAI MCP server. The runnable server implementation lives in the main Muntu backend.

The official Model Context Protocol (MCP) server for MuntuAI — giving AI agents programmatic access to email outreach campaigns, lead management, domain infrastructure, and real-time workspace events.

What This Is

MuntuAI is an AI-native email outreach platform. This MCP server lets external agents (Claude, Cursor, custom scripts) connect to a MuntuAI workspace and perform actions: create campaigns, import leads, verify domains, send emails, and subscribe to webhook events — all through a standard MCP interface over HTTP.

Protocol: Model Context Protocol over HTTP (JSON-RPC 2.0)
Transport: Streamable HTTP (text/event-stream or application/json)
Server protocol revision: 2025-03-26
Server URL: https://api.muntuai.com/api/mcp

Quick Start

1. Get an agent key

Log in to app.muntuai.com, go to Settings → Agent Keys, and create a key. Choose the autonomy level that matches what your agent needs:

You will see the raw key once: mnt_<keyId>.<secret>. Copy it immediately.

2. Configure your client

Security note: store the raw agent key in an environment variable or secret manager. Do not commit it to source control or paste it into shared files.

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "muntu": {
      "url": "https://api.muntuai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${env:MUNTU_AGENT_KEY}"
      }
    }
  }
}

Cursor (.cursor/mcp.json or Settings → MCP):

{
  "mcpServers": {
    "muntu": {
      "url": "https://api.muntuai.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ${env:MUNTU_AGENT_KEY}"
      }
    }
  }
}

Python:

pip install requests
export MUNTU_AGENT_KEY="mnt_YOUR_KEY_HERE"
python clients/python/muntu_mcp_client.py list-campaigns

Recommended:

export MUNTU_SESSION_ID="$(uuidgen)"

Muntu uses the optional MUNTU_SESSION_ID to group requests into one logical MCP session for session-scoped safety controls and budgeting.

3. Make your first call

Once connected, ask your agent:

"List all campaigns in my MuntuAI workspace"

The agent will call resources/read on muntu://campaigns/{workspaceId} and return the campaign list.

Or via Python CLI:

python clients/python/muntu_mcp_client.py list-campaigns
python clients/python/muntu_mcp_client.py list-domains
python clients/python/muntu_mcp_client.py list-emails

What Agents Can Do

Resources (read-only views)

Tools (30 total)

Grouped by domain — see tools/README.md for the full index.

• Campaigns — create, launch, pause, resume, monitor, review drafts, send
• Leads — import from array or CSV URL, enrich, sample
• Senders — create, delete, health check, assign to campaign
• **Dom

... View full README on GitHub