Perplexity MCP Server

This is a Python implementation of the Model Context Protocol (MCP) server for Perplexity. The server is built using FastAPI and follows the MCP specification.

Features

• Implements the core MCP endpoints:
  • /v1/models - List available models registered in the local MCP server
  • /v1/models/{model_id} - Get model information for a locally registered model
  • /v1/models/{model_id}/chat - Chat with a model - this endpoint actually calls the Perplexity API
  • /v1/models/{model_id}/complete - Text completion (demo implementation)
• Additional utility endpoints:
  • /perplexity-models - Lists the known models available in the Perplexity API
  • /api-key-test - Test if your Perplexity API key is properly configured
  • /server-info - Get information about the server
  • /health - Health check endpoint
• Ready for deployment

Running the Server

Locally

1. Create a .env file in the project root with your Perplexity API key:

   PERPLEXITY_API_KEY=your_actual_key_here
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Start the server:

   python main.py
   

The server will start on http://0.0.0.0:5000 (or the next available port).

Docker Deployment

1. Build the Docker image:

   docker build -t perplexity-mcp .
   

2. Run the container, passing your .env file:

   docker run --env-file .env -p 5003:5000 perplexity-mcp
   

   • The API will be available at http://localhost:5003
   • You can use any available port on your host (replace 5003 as needed)

Note:

• .env is included in .gitignore and .dockerignore to keep your API key secure.
• Never commit your .env file to GitHub.

Render Deployment

Render can automatically build and deploy this project from GitHub using the included Dockerfile.

1. Push your code to GitHub.
2. In the Render dashboard, create a new Web Service:
   • Select "Docker" as the environment.
   • Connect your GitHub repo and select the branch (e.g., main).
   • Set your environment variable PERPLEXITY_API_KEY in the Render dashboard (Settings > Environment).
   • Choose your desired port (default is 5000).
3. Deploy! Render will build and run your Docker container.

Security: Do not rely on .env in the repo for Render. Always set secrets in the Render dashboard.

API Documentation

Once the server is running, you can access the API documentation at /docs (e.g., http://0.0.0.0:5000/docs).

Model Registry

The server includes a simple in-memory model registry. In a production environment, you'd want to replace this with a more sophisticated registry system.

Adding Your Own Models

To add your own models, update the MODEL_REGISTRY dictionary in main.py with your model configurations.

Supported Models

The server supports the following Perplexity models:

• sonar - Perplexity's flagship model with strong reasoning
• sonar-small - Smaller, faster version of Sonar
• sonar-medium - Medium-sized version of Sonar
• sonar-pro - Pro version of Sonar with enhanced capabilities
• sonar-deep-research - Specialized for in-depth research tasks
• sonar-reasoning-pro - Advanced reasoning capabilities with enhanced logic
• codellama-70b - Specialized for code generation
• mixtral-8x7b - From Mistral AI, good for general tasks
• mistral-7b - Fast and efficient model from Mistral AI

To use a specific model, simply call /v1/models/{model_id}/chat with the model ID.

Supported Model Parameters

The chat endpoint supports the following parameters from the Perplexity API:

• max_tokens: Maximum number of tokens to generate
• temperature: Sampling temperature between 0 and 2
• top_p: Nucleus sampling parameter between 0 and 1
• top_k: Top-k sampling parameter
• presence_penalty: Presence penalty between -2 and 2
• frequency_penalty: Frequency penalty between -2 and 2
• stop: Stop sequences that cause the model to stop generating
• repetition_penalty: Repetition penalty for token generation
• logprobs: Whether to return log probabilities of the output tokens
• stream: Whether to stream the response