Bedrock Chat with MCP tool (ver.2)

April 6, 2025 ยท View on GitHub

This is a chat application built with Streamlit and integrated with the MCP (Model Context Protocol) tool.

Overview

Bedrock Chat with MCP tool is a chat application built with Streamlit and integrated with the MCP (Model Context Protocol) tool.

This application uses Bedrock . It interacts with the MCP (Model Context Protocol) server defined in config/mcp_config.json and accesses various tools. MCP is an open protocol that standardizes how applications provide context to LLM (https://modelcontextprotocol.io/). Chat history is stored in a YAML file.

The config/config.json file allows you to configure the LLM model to use, where the chat history files are stored, etc. The config/mcp_config.json file describes the configuration of the MCP server.

In the Streamlit sidebar, you can configure the following:

  • Select LLM model
  • Enable/disable prompt cache
  • Change chat history directory
  • Change MCP configuration file
  • Select tools
  • Generate system prompt
  • Start a new chat

Features

  • Chat interface using Streamlit
  • Bedrock integration
  • MCP tool integration
  • LLM model selection
  • Prompt cache enable/disable
  • Chat history save/load
  • MCP configuration file change
  • Tool selection
  • System prompt generation
  • File upload

Setup

  1. Install dependencies:

    uv sync

  2. Configure MCP server in config/mcp_config.json.

  3. Run the application.

    streamlit run main.py

Configuration

The config/config.json file is where you configure the LLM model and other settings.

{
    "chat_history_dir": "chat_history",
    "mcp_config_file": "config/mcp_config.json",
    "models": {
        "us.amazon.nova-pro-v1:0": {
            "cache_support": [
                "system"
            ]
        },
        ...
        "us.anthropic.claude-3-7-sonnet-20250219-v1:0": {
            "cache_support": [
                "system",
                "messages",
                "tools"
            ]
        },
        ...
    }
}

The config/mcp_config.json file contains the settings for the MCP server.

{
    "mcpServers": {
        "fetch": {
            "command": "docker",
            "args": [
                "run",
                "-i",
                "--rm",
                "mcp/fetch"
            ]
        },
        ...
    }
}

Usage

To run the Streamlit application, run the following command.

streamlit run main.py
  1. Run the Streamlit application.
  2. Enter a message in the chat input box.
  3. The chat model and MCP tool will generate a response.
  4. In the sidebar, you can configure the LLM model, chat history directory, MCP configuration file, etc.
  5. You can also select past chat history and resume the conversation.

Notes

  • Write the MCP server configuration in config/mcp_config.json.
  • To use Bedrock, you need an AWS account.
  • Chat history is stored in a YAML file.