Perplexity

MCP server for Perplexity AI integration with Claude Desktop - three complexity levels for different use cases

python

2

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "rohit-seelam-perplexity_mcp": {
      "command": "/Users/username/.local/bin/uv",
      "args": [
        "--directory",
        "/path/to/your/Perplexity_MCP",
        "run",
        "python",
        "server.py"
      ],
      "env": {
        "PERPLEXITY_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

You run a model context protocol (MCP) server that connects Perplexity AI’s search-enhanced language models with Claude Desktop. It provides three tools with increasing complexity to handle quick queries, technical analysis, and deep research, all while offering clean responses and robust debugging support for development workflows.

How to use

You use an MCP client to access the Perplexity tools from Claude Desktop. After you configure the MCP server, you can invoke the tools by name to perform tasks at the desired complexity level. The tools are designed to deliver structured responses with content and citations, while filtering internal thinking tokens.

How to install

Prerequisites: you need Python 3.11 or newer, the UV package manager, a Perplexity API key, and Claude Desktop installed on your machine.

Clone the Perplexity MCP project directory and navigate into it. You will work with a local MCP server script named server.py.
Install development or runtime dependencies using UV. If you are setting up for the first time, run the dependency synchronization command.
Create an environment file and set your API key for Perplexity. The environment variable PERPLEXITY_API_KEY must be present for the server to access the Perplexity API.
Start the MCP server locally to verify everything works. You should run the server in a way that matches your development setup and allows you to test the three tools.
Configure Claude Desktop to load the MCP server so you can use the tools in conversations.

Claude Desktop configuration snippet

{
  "mcpServers": {
    "perplexity-mcp": {
      "command": "/Users/username/.local/bin/uv",
      "args": [
        "--directory", 
        "/path/to/your/Perplexity_MCP",
        "run",
        "python",
        "server.py"
      ]
    }
  }
}

Environment and startup notes

Set the Perplexity API key in your environment so the MCP server can authenticate with Perplexity.

Tools you can use in Claude Desktop

The server exposes three tools, each with a distinct capability level for different use cases. You can call them by their names in Claude Desktop conversations to perform targeted tasks.

Available tools

perplexity_small

Fast queries with sonar-pro model for quick factual responses; aims for low latency (~3-10 seconds)

perplexity_medium

Enhanced reasoning with sonar-reasoning-pro for technical explanations; longer runtime (~10-30 seconds)

perplexity_large

Deep research with sonar-deep-research for comprehensive analysis; can take longer (up to ~30 minutes)