Mattermost MCP Host

What is Mattermost MCP Host?

Mattermost MCP Host is an integration that connects to Model Context Protocol (MCP) servers via a LangGraph‑based AI agent, enabling intelligent tool execution directly within Mattermost. It runs on Python 3.13.1+ and is built for teams that want to combine Mattermost chat with external services and AI‑driven automation.

How to use Mattermost MCP Host?

Clone the repository, install with uv sync, configure a .env file with your Mattermost server URL, bot token, and LLM API credentials, then define MCP servers in src/mattermost_mcp_host/mcp-servers.json. Start the integration by running mattermost-mcp-host. In Mattermost, simply chat with the bot in the configured channel or use the command prefix (default #) for direct operations like listing servers, tools, or calling a specific tool with JSON arguments.

Key features of Mattermost MCP Host

• 🤖 LangGraph agent integration for understanding and orchestrating user requests.
• 🔌 Connects to multiple MCP servers (stdio) defined in a JSON configuration.
• 🛠️ Dynamic tool loading – tools from MCP servers are automatically discovered and made available.
• 💬 Thread‑aware conversations maintain context within Mattermost threads.
• 🔄 Intelligent tool chaining – the agent can call multiple tools to fulfill a request.
• #️⃣ Direct command interface (#help, #servers, #<server> tools, etc.).

Use cases of Mattermost MCP Host

• Search GitHub issues and pull requests, then create a new issue if none exists – all from within a Mattermost channel.
• Use a web‑search MCP server (e.g., Tavily) to retrieve information and post results directly.
• Integrate custom MCP tools (e.g., file operations, database queries) and trigger them via chat or the # command prefix.

FAQ from Mattermost MCP Host

What does Mattermost MCP Host do?

It connects one or more MCP servers (each exposing tools, resources, and prompts) to Mattermost. A LangGraph agent interprets user messages, decides which MCP tools to call, and returns the response to the channel or thread.

Which AI providers are supported?

The default provider is Azure OpenAI, but the .env file can be configured for OpenAI, Anthropic Claude, or Google Gemini. See DEFAULT_PROVIDER and the corresponding API keys.

How do I add MCP servers?

Edit the file src/mattermost_mcp_host/mcp-servers.json with the command and arguments for each server (e.g., npx, uvx, docker). The integration connects via stdio and automatically discovers the tools each server offers.

What are the prerequisites?

Python 3.13.1+, the uv package manager, a running Mattermost instance, a Mattermost bot account with the required permissions, and access to an LLM API (Azure OpenAI by default). MCP servers must be installed and their dependencies available.

Is Mattermost MCP Host open‑source?

Yes, it is licensed under the MIT License. You can freely use, modify, and distribute it. The current version (0.1.0) is a beta release.