Label Studio MCP Server

What is Label Studio MCP Server?

Label Studio MCP Server is a Model Context Protocol (MCP) server that uses the official label-studio-sdk to allow programmatic management of Label Studio labeling projects, tasks, and predictions. It enables natural language or structured calls from MCP clients to interact with a Label Studio instance.

How to use Label Studio MCP Server?

Configure it by setting the LABEL_STUDIO_API_KEY and LABEL_STUDIO_URL environment variables. Add the server definition to your MCP client configuration (e.g., claude_desktop_config.json) using the uvx command and specify the required environment variables. Then invoke tools like create_label_studio_project_tool, import_label_studio_project_tasks_tool, and create_label_studio_prediction_tool through your MCP client.

Key features of Label Studio MCP Server

• Project management: Create, update, list, and view project details/configurations.
• Task management: Import tasks from files, list tasks, retrieve data and annotations.
• Prediction integration: Add model predictions to specific tasks.
• SDK integration: Leverages official label-studio-sdk for reliable communication.
• Natural language interaction: Query project status and task counts via MCP clients.

Use cases of Label Studio MCP Server

• Automated project setup: Create labeling projects with custom XML configs programmatically.
• Bulk task import: Import thousands of tasks from a JSON file into a project in one step.
• ML prediction injection: Add predictions from machine learning models to existing tasks.
• Project status queries: Ask natural language questions like “How many tasks are labeled in my RAG review project?”.
• Template updates: Update labeling configurations without manual UI navigation.

FAQ from Label Studio MCP Server

What dependencies are required?

A running Label Studio instance accessible from the server, and an API key from your Label Studio user account. The server uses Python, uv, and the label-studio-sdk.

How do I configure authentication?

Set the LABEL_STUDIO_API_KEY environment variable to your Label Studio user account API key, and LABEL_STUDIO_URL to the URL of your Label Studio instance.

Where does my data live?

All data remains in your Label Studio instance. The server only reads from and writes to that instance via its API; it does not store data locally.

Are there any known limits?

Listing tasks returns up to 100 task IDs per call. The task import tool expects a JSON file containing a list of task objects. Predictions must follow the Label Studio format as a list of dictionaries.

What transport does the MCP server use?

The server is launched as a subprocess by the MCP client (e.g., via claude_desktop_config.json) and communicates over standard input/output using the MCP protocol.