MQTT MCP Server
A lightweight Model Context Protocol (MCP) server that connects LLM agents to MQTT devices in a secure, standardized way, enabling seamless integration of AI-driven workflows with Building Automation (BAS), Industrial Control (ICS) and Smart Home systems, allowing agents to monitor real-time sensor data, actuate devices, and orchestrate complex automation tasks.
Getting Started
The server is built with FastMCP 2.0 and uses uv for project and dependency management. Simply run the following command to install uv or check out the installation guide for more details and alternative installation methods.
curl -LsSf https://astral.sh/uv/install.sh | sh
Clone the repository, then use uv to install project dependencies and create a virtual environment.
git clone https://github.com/ezhuk/mqtt-mcp.git
cd mqtt-mcp
uv sync
Start the MQTT MCP server by running the following command in your terminal. It defaults to using the Streamable HTTP transport on port 8000.
uv run mqtt-mcp
To confirm the server is up and running and explore available resources and tools, run the MCP Inspector and connect it to the MQTT MCP server at http://127.0.0.1:8000/mcp/. Make sure to set the transport to Streamable HTTP.
npx @modelcontextprotocol/inspector
Core Concepts
The MQTT MCP server leverages FastMCP 2.0's core building blocks - resource templates, tools, and prompts - to streamline MQTT receive and publish operations with minimal boilerplate and a clean, Pythonic interface.
Receive Message
Each topic on a device is mapped to a resource (and exposed as a tool) and resource templates are used to specify connection details (host, port) and receive parameters (topic, timeout).
@mcp.resource("mqtt://{host}:{port}/{topic*}")
@mcp.tool(
annotations={
"title": "Receive Message",
"readOnlyHint": True,
"openWorldHint": True,
}
)
async def receive_message(
topic: str,
host: str = settings.mqtt.host,
port: int = settings.mqtt.port,
timeout: int = 60,
) -> str:
"""Receives a message published to the specified topic, if any."""
...
Publish Message
Publish operations are exposed as a tool, accepting the same connection details (host, port) and allowing to publish a message to a specific topic in a single, atomic call.
@mcp.tool(
annotations={
"title": "Publish Message",
"readOnlyHint": False,
"openWorldHint": True,
}
)
async def publish_message(
topic: str,
message: str,
host: str = settings.mqtt.host,
port: int = settings.mqtt.port,
) -> str:
"""Publishes a message to the specified topic."""
...
Authentication
To enable Bearer Token authentication for the Streamable HTTP transport, provide the RSA public key in PEM format in the .env file. Check out the Bearer Token Authentication section for more details.
Interactive Prompts
Structured response messages are implemented using prompts that help guide the interaction, clarify missing parameters, and handle errors gracefully.
@mcp.prompt(name="mqtt_help", tags={"mqtt", "help"})
def mqtt_help() -> list[Message]:
"""Provides examples of how to use the MQTT MCP server."""
...
Here are some example text inputs that can be used to interact with the server.
Publish {"foo":"bar"} to topic "devices/foo" on 127.0.0.1:1883.
Receive a message from topic "devices/bar", waiting up to 30 seconds.
Examples
The examples folder contains sample projects showing how to integrate with the MQTT MCP server using various client APIs to provide tools and context to LLMs.
- openai-agents - shows how to connect to the MQTT MCP server using the OpenAI Agents SDK.
- openai - a minimal app leveraging remote MCP server support in the OpenAI Python library.
Docker
The MQTT MCP server can be deployed as a Docker container. Build the image using the docker build command:
docker build -t mqtt-mcp .
Then run the container as follows:
docker run -dit \
--name mqtt-mcp \
--restart=always \
-p 8080:8000 \
--env-file .env \
mqtt-mcp
This maps port 8080 on the host to the MCP server's port 8000 inside the container and loads settings from the .env file, if present.
License
The server is licensed under the MIT License.