The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the MuntuAI listing page.
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.
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
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:
| Level | What it can do |
|---|---|
observer | Read anything — campaigns, leads, domains, senders, events |
copilot | Read + create and modify — campaigns, domains, senders, leads, webhooks |
autonomous | Everything above + pause, resume, delete |
You will see the raw key once: mnt_<keyId>.<secret>. Copy it immediately.
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):
Cursor (.cursor/mcp.json or Settings → MCP):
Python:
Recommended:
Muntu uses the optional MUNTU_SESSION_ID to group requests into one logical MCP session for session-scoped safety controls and budgeting.
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:
| Resource | URI | Description |
|---|---|---|
| workspace | muntu://workspace/{id} | Profile, policy, infrastructure counts |
| domains | muntu://domains/{id}?cursor=0&limit=50 | All domains with DNS and verification state |
| senders | muntu://senders/{id}?cursor=0&limit=50 | All email accounts with status |
| campaigns | muntu://campaigns/{id}?cursor=0&limit=20 | Unified campaign list |
| campaign-plans | muntu://campaign-plans/{id}?cursor=0&limit=20 | Draft review queue |
| events | muntu://events/{id}?cursor=0 | Recent system events (filterable) |
Grouped by domain — see tools/README.md for the full index.
Workspace — All data in MuntuAI is scoped to a workspace. Your agent key is bound to one workspace and can only access data within it.
Autonomy levels — Every tool has a minimum required level. If your key's level is below the requirement, the call is rejected and the attempt is logged. See docs/autonomy-levels.md.
Campaign lifecycle — uploaded → enriching → drafting → sending → completed. Agents can observe this lifecycle through get_campaign_performance and the events resource. See workflows/monitor-campaign.md.
Domain verification — Before sending, a domain must be verified with the email provider. This involves publishing DNS records and triggering verification. See workflows/domain-setup.md.