Robust MCP Client

A versatile Chat Client that supports the Model Context Protocol (MCP), works via CLI and Web UI, and integrates with both local and cloud LLMs.

Features

• Dual Interface: CLI and Web UI (React + Tailwind).
• Local LLM: Built-in support for llama-cpp-python (auto-downloads TinyLlama).
• Cloud LLM: Support for OpenAI, Gemini, and Claude.
• MCP Support: Connect to multiple MCP servers via mcp.json.
• Memory: Persistent conversation history (SQLite).
• Rich UI: Streaming responses, Markdown, Code Highlighting, Mermaid Diagrams.

Setup

1. Install Dependencies:

   pip install -r requirements.txt
   cd ui && npm install

2. Configuration:

   • Create a .env file (optional) for API keys:

     OPENAI_API_KEY=...
     GEMINI_API_KEY=...
     ANTHROPIC_API_KEY=...
     

   • Edit mcp.json to add MCP servers.

Usage

CLI

Run the chat interface in the terminal:

python -m cli.main chat

Web UI

1. Start the backend server:

   python -m cli.main serve

2. Start the frontend (in a separate terminal):

   cd ui
   npm run dev

3. Open http://localhost:5173

Docker

Quick Start (Ephemeral)

To try it out quickly without persisting data:

docker run -p 8000:8000 alexmerced/robust_mcp_client

Note: Models will be re-downloaded and chat history lost when the container stops.

Using Pre-built Image (Persistent)

To persist models and history:

docker run -p 8000:8000 \
  -v $(pwd)/models:/app/models \
  -v $(pwd)/history.db:/app/history.db \
  alexmerced/robust_mcp_client

Building Locally

You can also build the image yourself:

1. Build the image:

   docker build -t robust-mcp-client .

2. Run the container:

   docker run -p 8000:8000 \
     -v $(pwd)/models:/app/models \
     -v $(pwd)/history.db:/app/history.db \
     robust-mcp-client

   • -p 8000:8000: Expose the server port.
   • -v .../models: Persist downloaded models.
   • -v .../history.db: Persist conversation history.

3. Access the UI: Open http://localhost:8000

Documentation

• Architecture
• Configuration
• Development
• Usage
• API Reference

Development

• Run tests: python -m pytest tests/