Label Studio

You can manage a Label Studio instance programmatically through a dedicated MCP server. This server lets you create and configure projects, import and inspect tasks, and add model predictions directly from MCP clients using the official label-studio-sdk.

How to use

Connect to the Label Studio MCP Server from your MCP client and start performing operations like creating projects, importing tasks, listing tasks, and attaching predictions. You can manage labeling configurations, access task data and annotations, and push model results into specific tasks. All actions are exposed through the server’s project, task, and prediction tools, enabling end-to-end automation for labeling workflows.

How to install

Prerequisites you need before starting:

• Yes, running Label Studio is required

• Obtainable from user account settings

Step-by-step setup

1. Prepare the MCP server configuration so it can reach your Label Studio instance and use your API key.

2. Create an MCP server entry that uses a local runtime command to start the server. The following configuration runs the MCP server and injects the Label Studio connection details via environment variables.

{
  "mcpServers": {
    "label_studio": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/HumanSignal/label-studio-mcp-server",
        "mcp-label-studio"
      ],
      "env": {
        "LABEL_STUDIO_API_KEY": "your_actual_api_key_here",
        "LABEL_STUDIO_URL": "http://localhost:8080"
      }
    }
  }
}

What to run and where

Run the MCP server using the configuration file that contains the mcpServers entry shown above. This starts the MCP server in a runtime environment that can communicate with Label Studio through the API key and URL you provide.

Notes on environment and startup

Ensure the Label Studio URL is reachable from the MCP server host and that the API key is kept secret. If you deploy to a client-managed environment, keep your configuration in a secure location and reference it from your MCP client.

Additional sections

Configuration details, security considerations, and practical usage tips are provided below to help you tailor the MCP server to your labeling workflow.

Configuration

You configure the MCP server with the following entry to connect to Label Studio. This entry uses a local runtime to launch the MCP server and passes the API key and URL through environment variables.

{
  "mcpServers": {
    "label_studio": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/HumanSignal/label-studio-mcp-server",
        "mcp-label-studio"
      ],
      "env": {
        "LABEL_STUDIO_API_KEY": "your_actual_api_key_here",
        "LABEL_STUDIO_URL": "http://localhost:8080"
      }
    }
  }
}

Security and access

Protect your API key and restrict access to the MCP endpoint. Use network controls and rotate API keys on a regular basis to minimize risk.

Examples and common tasks

Create a project, import tasks from a JSON file, view task data, and attach predictions to tasks using the available tools described in the Tools section.

Notes

This MCP server uses the official label-studio-sdk to communicate with Label Studio and provides a set of tools to manage projects, tasks, and predictions in an automated fashion.

Available tools

get_label_studio_projects_tool

Lists available Label Studio projects with their IDs, titles, and task counts.

get_label_studio_project_details_tool

Fetches detailed information for a specific project by ID.

get_label_studio_project_config_tool

Retrieves the XML labeling configuration for a project by ID.

create_label_studio_project_tool

Creates a new Label Studio project with a title, label config, and optional settings; returns project details including a URL.

update_label_studio_project_config_tool

Updates the labeling XML configuration for an existing project.

list_label_studio_project_tasks_tool

Lists up to 100 task IDs within a project.

get_label_studio_task_data_tool

Retrieves the data payload for a specific task.

get_label_studio_task_annotations_tool

Fetches existing annotations for a specific task.

import_label_studio_project_tasks_tool

Imports tasks from a JSON file into a project and returns an import summary.

create_label_studio_prediction_tool

Adds a prediction to a specific task with a structured result.