README.md

---
title: Rag Mcp Server
emoji: 🏢
colorFrom: indigo
colorTo: pink
sdk: gradio
sdk_version: 5.31.0
app_file: app.py
pinned: false
short_description: 'This space defines a RAG MCP server '
---

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

# Rag MCP server

This is a simple gradio server that allows you to run a RAG (Retrieval-Augmented Generation) model using the
MCP (Model Control Protocol) interface.

# Requirements

- UV: The python package manager. Visit https://github.com/astral-sh/uv

Since the server uses SERPER API, you will need to set the `SERPER_API_KEY` environment variable to
enable the search functionality. You can export it in your shell:

```bash
export SERPER_API_KEY=your_serper_api_key
```

# Installation

```bash
uv sync
```

# Usage

Just run the gradio application:

```bash
uv run gradio app.py
```

# Access the server

Open your browser and go to `http://localhost:7860` to access the RAG MCP server.

You can see the [gradio docs](https://www.gradio.app/guides/building-mcp-server-with-gradio) for more information on how
to use the interface.

# Using the MCP server with a mpc client

You can use the tiny-agents client to interact with the MCP server. To do this, you need to install the huggingface_hub
package:

```bash
pip install huggingface_hub
```

Then, you can use the following code to interact with the MCP server:

```bash
export HF_TOKEN=your_huggingface_token

tiny_agents run agent.json
```

The `agent.json` file contains the configuration for the MCP client. Here is an example of how to create this file:

```json
{
  "model": "meta-llama/Llama-3.3-70B-Instruct",
  "provider": "cerebras",
  "servers": [
    {
      "type": "sse",
      "config": {
        "url": "https://huggingface.co/proxy/frascuchon-rag-mcp-server.hf.space/gradio_api/mcp/sse"
      }
    }
  ]
}
```

You can change the server URL to point to your own MCP server if you are running it locally or on a different host.