A robust Python client for interacting with Model Context Protocol (MCP) servers. Built with LangChain and LangGraph, it enables you to chat with multiple MCP servers simultaneously using a natural language interface on the CLI or a Web UI.

• Multi-Server Support: Connect to unlimited MCP servers (local processes or remote HTTP/SSE).
• Secure Authentication: Securely inject API keys and OAuth tokens via Environment Variables (${VAR}).
• Streaming Responses: Real-time token-by-token generation in the CLI.
• Customizable Agent: Configure the System Prompt and LLM Provider via .env.
• Web Interface: Optional Flask-based UI for browser interaction.
• Dremio Integration: Native helper script for Dremio OAuth flow.

Clone the repository and install dependencies:

git clone https://github.com/AlexMercedCoder/langchain-mcp-client.git
cd langchain-mcp-client
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Create a .env file from the example:

cp .env.example .env

Edit .env to add your LLM API Key (e.g., OpenAI):

OPENAI_API_KEY=sk-...
LLM_MODEL=openai:gpt-4o
SYSTEM_PROMPT="You are a helpful assistant."

Start the interactive CLI:

python client.py

Define your servers in mcp_servers.json.

Connect to local processes using the stdio transport. Ideal for secure, local-only tools like filesystem access or SQLite interaction.

{
  "filesystem": {
    "command": "npx",
    "args": ["-y", "@modelcontextprotocol/server-filesystem", "/home/user/files"],
    "transport": "stdio"
  },
  "sqlite": {
    "command": "uvx",
    "args": ["mcp-server-sqlite", "--db-path", "./my.db"],
    "transport": "stdio"
  }
}

Connect to public MCP servers via HTTP.

{
  "weather": {
    "url": "https://weather-mcp.example.com",
    "transport": "streamable_http"
  }
}

Never commit secrets to mcp_servers.json. Instead, use the ${VAR_NAME} syntax. The client will automatically substitute these with values from your .env file at runtime.

In .env:

MY_OAUTH_TOKEN=secret-123

In mcp_servers.json:

{
  "secure-api": {
    "url": "https://api.example.com/mcp",
    "transport": "streamable_http",
    "headers": {
      "Authorization": "Bearer ${MY_OAUTH_TOKEN}"
    }
  }
}

To query Dremio Data Catalogs, you must authenticate via OAuth.

1. Create App: In Dremio, register a "Native" OAuth App with Redirect URI: http://localhost:8000/callback
2. Configure .env:

   DREMIO_CLIENT_ID=your-client-id

3. Get Token: Run the helper script:

   python auth_dremio.py

   Login via the browser link provided.
4. Save Token: Copy the resulting token to .env:

   DREMIO_TOKEN=eyJhbGci...

5. Configure Server:

   "dremio": {
     "url": "https://mcp.dremio.cloud/mcp/<PROJECT_ID>",
     "transport": "streamable_http",
     "headers": {
       "Authorization": "Bearer ${DREMIO_TOKEN}"
     }
   }

Customize the agent's personality or rules by setting SYSTEM_PROMPT in .env.

SYSTEM_PROMPT="You are a SQL expert. Always check schemas before querying."

Prefer a browser UI?

python webui.py

Open http://localhost:5000 to chat.

The client uses LangChain, so it supports OpenAI, Anthropic, Google Vertex, etc. Change LLM_MODEL in .env:

• openai:gpt-4o
• anthropic:claude-3-sonnet (requires ANTHROPIC_API_KEY)
• google:gemini-pro (requires GOOGLE_API_KEY)

Keep the core libraries up to date:

pip install --upgrade -r requirements.txt

Run tests to ensure environment substitution and agent creation are working:

python verify_changes.py

MIT