MCP Server

The ThingsBoard MCP Server connects your MCP-enabled LLM client to a live ThingsBoard instance via the Model Context Protocol. It exposes a comprehensive set of tools across multiple tool groups, letting you query, create, and manage platform data using natural language — no REST calls or dashboard clicks required.

Capabilities at a glance:

• Devices & assets — create, search, update, and delete entities
• Telemetry — read and write attributes and time-series data, with aggregation and TTL support
• Alarms — create, acknowledge, clear, and query alarms by severity
• Relations — create and navigate entity relationships
• OTA packages — upload, download, and assign firmware/software to devices
• Entity data queries — run complex filtered queries with attribute and telemetry filters
• Customers & users — full CRUD and user management

Works with Claude Desktop, Cursor, VS Code Copilot, Claude Code, and any MCP-compatible client.

Before you begin, ensure that you have:

• A ThingsBoard Cloud account or an on-premises ThingsBoard instance.
• Authentication credentials — either an API key (recommended, requires ThingsBoard 4.3+) or username and password.
• An MCP-enabled LLM client — for example, Claude Desktop, Claude Code, Cursor, VS Code with an MCP extension, or any other client that supports the Model Context Protocol.

Pull the Docker image:

docker pull thingsboard/mcp

Choose your MCP client below. Each tab shows the configuration for API key authentication (recommended). For username/password authentication, see the Username & Password section below.

{
  "mcpServers": {
    "thingsboard": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "THINGSBOARD_URL",
        "-e", "THINGSBOARD_API_KEY",
        "thingsboard/mcp"
      ],
      "env": {
        "THINGSBOARD_URL": "<thingsboard_url>",
        "THINGSBOARD_API_KEY": "<your_api_key>"
      }
    }
  }
}

{
  "mcpServers": {
    "thingsboard": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "THINGSBOARD_URL",
        "-e", "THINGSBOARD_API_KEY",
        "thingsboard/mcp"
      ],
      "env": {
        "THINGSBOARD_URL": "<thingsboard_url>",
        "THINGSBOARD_API_KEY": "<your_api_key>"
      }
    }
  }
}

{
  "mcp": {
    "servers": {
      "thingsboard": {
        "command": "docker",
        "args": [
          "run", "-i", "--rm",
          "-e", "THINGSBOARD_URL",
          "-e", "THINGSBOARD_API_KEY",
          "thingsboard/mcp"
        ],
        "env": {
          "THINGSBOARD_URL": "<thingsboard_url>",
          "THINGSBOARD_API_KEY": "<your_api_key>"
        }
      }
    }
  }
}

claude mcp add thingsboard -e THINGSBOARD_URL=<thingsboard_url> -e THINGSBOARD_API_KEY=<your_api_key> -- docker run -i --rm -e THINGSBOARD_URL -e THINGSBOARD_API_KEY thingsboard/mcp

docker run --rm -p 8000:8000 -e THINGSBOARD_URL=<thingsboard_url> -e THINGSBOARD_API_KEY=<your_api_key> -e SPRING_AI_MCP_SERVER_STDIO=false -e SPRING_WEB_APPLICATION_TYPE=servlet thingsboard/mcp

If you cannot use an API key (e.g., on ThingsBoard versions before 4.3), replace the API key environment variables with username and password. The example below shows Claude Desktop config — the same args and env pattern applies to all other clients.

{
  "mcpServers": {
    "thingsboard": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "THINGSBOARD_URL",
        "-e", "THINGSBOARD_USERNAME",
        "-e", "THINGSBOARD_PASSWORD",
        "thingsboard/mcp"
      ],
      "env": {
        "THINGSBOARD_URL": "<thingsboard_url>",
        "THINGSBOARD_USERNAME": "<thingsboard_username>",
        "THINGSBOARD_PASSWORD": "<thingsboard_password>"
      }
    }
  }
}

Once your client starts, you will see thingsboard tools available:

Here are some prompts to try once the MCP server is connected:

Devices — “Get all devices of type ‘Temperature Sensor’ and show their latest telemetry.”

Telemetry — “Generate sample temperature data for Sensor A1 for the last 7 days with a 30-minute sampling rate.”

Alarms — “Show all active alarms with CRITICAL severity and who they are assigned to.”

Anomaly detection — “Analyze the power consumption of Building 3 for the past week and flag any anomalies.”

Relations — “Show all devices related to asset ‘Factory Floor 1’ and their connection types.”

Entity queries — “Find all devices where the ‘firmware_version’ attribute is older than 2.0 and battery level is below 20%.”

OTA updates — “List all devices with outdated firmware and assign the latest OTA package to them.”

Here is a list of devices inside the ThingsBoard platform:

Using your MCP client, make requests in natural language:

Get devices of type 'Air Quality Sensor'

ThingsBoard MCP translates this into a ThingsBoard API query and returns structured results:

MCP can simulate (generate) telemetry for any device based on your request.

Generate sample data for Office #1 AQI Sensor for the last 3 days and save it as timeseries data. Use sampling rate as 1 hour

Requesting sample data generation via MCP. MCP automatically adds the generated time series into ThingsBoard.

Once telemetry is available, request MCP to analyze it:

Analyze AQI of Office #1 AQI Sensor for anomalies in the last 3 days and show results in a table.

Requesting anomaly analysis via MCP. MCP returns anomaly detection results in a tabular format.

The MCP server provides tools organized into the following groups:

See the ThingsBoard MCP GitHub repository for a full tool reference.

Each tool group can be individually enabled or disabled via environment variables. All are enabled by default.

For example, to enable only device and telemetry tools, set every other group to false:

{
  "mcpServers": {
    "thingsboard": {
      "command": "docker",
      "args": [
        "run", "-i", "--rm",
        "-e", "THINGSBOARD_URL",
        "-e", "THINGSBOARD_API_KEY",
        "-e", "THINGSBOARD_TOOLS_ASSET",
        "-e", "THINGSBOARD_TOOLS_CUSTOMER",
        "-e", "THINGSBOARD_TOOLS_USER",
        "-e", "THINGSBOARD_TOOLS_ALARM",
        "-e", "THINGSBOARD_TOOLS_RELATION",
        "-e", "THINGSBOARD_TOOLS_OTA",
        "-e", "THINGSBOARD_TOOLS_EDQ",
        "thingsboard/mcp"
      ],
      "env": {
        "THINGSBOARD_URL": "<thingsboard_url>",
        "THINGSBOARD_API_KEY": "<your_api_key>",

In SSE mode, the server listens on http://<HTTP_BIND_ADDRESS>:<HTTP_BIND_PORT>/sse (default: http://127.0.0.1:8000/sse).

The ThingsBoard MCP Server extends beyond AQI anomaly detection. It applies to a wide range of scenarios:

By combining MCP’s natural language interface with ThingsBoard’s data visualization and alerting capabilities, you can rapidly prototype and deploy AI-driven IoT solutions across many industries.