Zammad MCP Server: Open Source Setup for Claude and Cursor

Install and use the open-source Zammad MCP Server for Claude, Cursor, and MCP clients with typed ticket tools, safe access control, and Zammad API token setup.

Zammad MCP Server

The Zammad MCP Server is a free, open-source Model Context Protocol (MCP) server for Zammad. It connects Claude Desktop, Cursor, and other MCP clients to your Zammad helpdesk so AI agents can search tickets, summarize conversations, draft replies, and inspect users, organizations, groups, and system metadata through typed tools.

This page is the main documentation hub for the Zammad MCP server. Use it if you want to install zammad-mcp-server, connect it to Claude or Cursor, compare it with Zammad’s native AI features, or understand whether MCP is the right integration layer for your helpdesk automation.

• License: MIT
• Package: zammad-mcp-server on PyPI
• Source: github.com/Softoft-Orga/zammad-mcp-server
• Primary use case: external AI assistants working with Zammad via the Zammad REST API
• Zammad compatibility: 6.0+ via REST API — see matrix below

Zammad version compatibility

The MCP package (zammad-mcp-server) is versioned independently from Zammad. One PyPI release works with every row below; point ZAMMAD_URL at your instance.

Local dev stack: clone zammad-mcp-server/docker, copy .env.example → .env, and set ZAMMAD_VERSION (7.1, 7.0, 6.5, or 6.3). Full details: COMPATIBILITY.md on GitHub.

Use get_server_info after health_check to confirm the connected Zammad version.

Who is this for?

Install the Zammad MCP Server

For most local MCP clients, the fastest path is uvx:

uvx zammad-mcp-server

You can also install it permanently:

pip install zammad-mcp-server
uv tool install zammad-mcp-server

The server needs a Zammad URL and a personal access token. Start with the Quick Start for screenshots of the Zammad token flow, then add the server to Claude Desktop or Cursor.

How this fits the Zammad + AI landscape

flowchart LR
  subgraph native [Zammad 7.1 native AI]
    Z7[Summaries and writing assistant inside Zammad UI]
  end
  subgraph mcp [Zammad MCP Server]
    MCP[Claude Desktop Cursor custom agents]
  end
  subgraph otai [Open Ticket AI Runtime]
    RT[Custom trained models on-prem routing]
  end
  Zammad[(Zammad REST API)]
  Z7 --> Zammad
  MCP --> Zammad
  RT --> Zammad

• Zammad 7.1 native AI — built-in summaries, writing assistant, and AI agents inside the Zammad UI.
• Zammad MCP Server (this project) — connects external MCP clients (Claude Desktop, Cursor, your own agents) to Zammad.
• Open Ticket AI Runtime — on-prem inference with custom-trained classification models via the OTAI Zammad connector plugin (see Zammad + OTAI integration).

All three can coexist. MCP is the fastest path to “talk to my helpdesk from Claude.”

What you can do with Zammad MCP

The Zammad MCP server exposes Zammad as structured tools instead of raw REST calls. Common workflows include:

Key features

• 30+ MCP tools for tickets, articles, users, organizations, and groups
• Access control via environment variables (category allow/deny, tool deny-list, group scope)
• Multiple auth modes — API token (recommended), OAuth2, basic auth
• Install options — pip, uvx, Docker, SSE transport for remote clients
• Built-in prompts — ticket summary, customer communication, escalation analysis
• Resource URIs — zammad://ticket/{id}, zammad://user/{id}, zammad://states

Documentation map

Zammad MCP Server vs alternatives

Choose the Zammad MCP Server when the search intent is “connect my AI assistant to Zammad now.” Choose Open Ticket AI Runtime when you need autonomous queue routing, classification, or custom models trained on your historical ticket data.

FAQ

Is the Zammad MCP Server open source?

Yes. The server is MIT licensed and published as zammad-mcp-server on PyPI with source code on GitHub.

Does it work with Claude Desktop and Cursor?

Yes. Both Claude Desktop and Cursor can launch the Zammad MCP Server with uvx zammad-mcp-server and pass ZAMMAD_URL plus ZAMMAD_HTTP_TOKEN through the MCP config.

Do I need admin rights in Zammad?

Not for every workflow. Most support workflows need a personal access token with ticket access. Admin-level tools require matching Zammad permissions and should be scoped carefully.

Can the MCP server delete tickets?

Destructive tools can be denied through MCP_DENIED_TOOLS. For production and shared assistant setups, deny delete operations unless you have a strong review process.

Which Zammad versions are supported?

See the compatibility matrix on GitHub. In short: 6.0+ is supported; 7.0 and 7.1 are tested. The bundled dev stack defaults to 7.1 and accepts ZAMMAD_VERSION in docker/.env for 7.0, 6.5, or 6.3.

Is this the same as Zammad’s native AI features?

No. Zammad native AI runs inside the Zammad interface. The Zammad MCP Server is an integration layer for external MCP clients such as Claude, Cursor, and custom agents.

Learn more

• Tutorial: Zammad MCP Server — setup and usage (blog)
• Repository: Softoft-Orga/zammad-mcp-server
• Package: zammad-mcp-server on PyPI

Next steps

1. Follow the Quick Start guide.
2. Harden access with Security before production use.
3. Need custom models trained on your queues? Explore Open Ticket AI for Zammad.