Perplexity MCP Server

A lightweight local MCP server for querying Perplexity AI. This server provides a simple interface to ask questions and get answers from Perplexity's API.

Disclaimer

This repository was generated with Claude Sonnet 4.5

Setup

Option 1: Local Installation

1. Install dependencies:

   pip install -r requirements.txt

2. Set your API key:

   export PERPLEXITY_API_KEY="your-api-key-here"

Option 2: Docker Container

1. Build the image:

   docker build -t perplexity-mcp .

2. Set your API key:

   export PERPLEXITY_API_KEY="your-api-key-here"

3. Run with docker (recommended):

   docker run -it \
     -e PERPLEXITY_API_KEY="${PERPLEXITY_API_KEY} \
     perplexity-mcp

Usage

Running the Server

STDIO Transport (default):

python perplexity_mcp.py

The server will start and wait for requests over stdin/stdout.

HTTP Transport:

python perplexity_mcp.py --http

The server will start with HTTP transport on 0.0.0.0:8000 (default). You can customize the host and port using environment variables:

HOST=127.0.0.1 PORT=9000 python perplexity_mcp.py --http

HTTP Transport with Docker:

docker run -p 8000:8000 \
  -e PERPLEXITY_API_KEY="${PERPLEXITY_API_KEY}" \
  perplexity-mcp \
  python perplexity_mcp.py --http

Custom host/port:

docker run -p 9000:9000 \
  -e PERPLEXITY_API_KEY="${PERPLEXITY_API_KEY}" \
  -e HOST=0.0.0.0 \
  -e PORT=9000 \
  perplexity-mcp \
  python perplexity_mcp.py --http

Tools Available

ask_perplexity

Ask a question to Perplexity AI and get an answer.

Parameters:

• question (string, required): The question to ask Perplexity AI
• model (string, optional): Which model to use
  • "sonar" (default): Faster responses
  • "sonar-pro": More detailed/comprehensive responses

Example requests:

Quick question with default model:

{
  "question": "What is quantum entanglement?"
}

Detailed question with pro model:

{
  "question": "Explain the history and current state of quantum computing",
  "model": "sonar-pro"
}

Integrating with Claude

Local Installation

Add it to your claude_desktop_config.json:

{
  "mcpServers": {
    "perplexity": {
      "command": "python",
      "args": ["/path/to/perplexity_mcp.py"],
      "env": {
        "PERPLEXITY_API_KEY": "${PERPLEXITY_API_KEY}"
      }
    }
  }
}

Docker Container

{
  "mcpServers": {
    "perplexity": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "PERPLEXITY_API_KEY=${PERPLEXITY_API_KEY}",
        "perplexity-mcp"
      ]
    }
  }
}

Make sure to set PERPLEXITY_API_KEY in your shell environment.

Integrating with other tools

Check the docs for that specific tool, the MCP setup will be similar to Claude Code.

Troubleshooting

Error: PERPLEXITY_API_KEY environment variable is not set

• Make sure you've exported the environment variable: export PERPLEXITY_API_KEY="your-key" or use a .env file for environment config.

Error: Authentication failed

• Check that your API key is correct and has not expired

Error: Rate limit exceeded

• Wait a moment before making another request. Perplexity has rate limits on API usage.

Timeout errors

• The Perplexity API took too long to respond. Try again or use the faster "sonar" model instead of "sonar-pro".

Notes

• This is a simple server designed for personal use
• The server uses async/await for non-blocking I/O
• Responses are limited by the Perplexity API's capabilities and rate limits
• The server supports both stdio and HTTP (streamable-http) transports