Gradio 5.38.0 MCP Server Improvements

Gradio version 5.38.0 introduces five major enhancements to its Model Context Protocol (MCP) server implementation, streamlining how developers build, host, and integrate AI-powered tools on Hugging Face Spaces.

Seamless Local File Support

Remote Gradio MCP servers previously required input files (such as images, video, or audio) to be accessible via public URLs, as the server and client were hosted on different machines. To eliminate the manual step of hosting files online, Gradio now includes a "File Upload" MCP server.

This dedicated server allows LLM agents to upload files directly to a Gradio application. When tools in a Gradio MCP server require file inputs, the connection documentation now automatically provides instructions on how to initialize the "File Upload" MCP server to facilitate this process.

Real-time Progress Notifications

To improve user experience during long-running AI tasks, Gradio now streams progress notifications to the MCP client. This allows users to monitor the execution status of a tool in real-time. Developers are encouraged to implement their MCP tools to emit these progress statuses to leverage this functionality.

Automated OpenAPI Specification Transformation

Gradio has introduced the gr.load_openapi function to automate the integration of existing RESTful APIs into LLM environments. Previously, developers had to manually map API endpoints to MCP tools, a process that was often time-consuming and error-prone.

With gr.load_openapi, developers can create a Gradio application directly from an OpenAPI schema (a machine-readable JSON file). By launching the application with mcp_server=True, Gradio automatically transforms the OpenAPI specification into an MCP server, enabling a business backend to be integrated into any MCP-compatible LLM with a single line of code.

Enhanced Authentication via gr.Header

To simplify the management of authentication headers when calling services on behalf of users, Gradio now supports the gr.Header type for MCP server arguments.

When a developer types an argument as gr.Header, Gradio automatically extracts the corresponding header from the incoming request and passes it to the function. Additionally, the MCP connection documentation automatically displays the required headers, clearly communicating to users which credentials must be supplied to connect to the server.

Customizable Tool Descriptions

While Gradio automatically generates tool descriptions from function names and docstrings, developers can now explicitly override these descriptions using the api_description parameter. This allows for more precise control over how a tool is described to the LLM, ensuring the agent understands the tool's purpose and application more accurately.