Perplexity Search MCP logo

Perplexity Search MCP

by arjunkmrm

A simple Model Context Protocol (MCP) server for Perplexity's web search using the sonar or sonar-pro models. It provides a tool for AI assistants to perform web searches powered by Perplexity's API.

View on GitHub

Last updated: N/A

What is Perplexity Search MCP?

This server is a Model Context Protocol (MCP) server designed to enable AI assistants to perform web searches using Perplexity's chat completions API with the sonar or sonar-pro models.

How to use Perplexity Search MCP?

To use this server, you need to set the PERPLEXITY_API_KEY environment variable with your Perplexity API key. Then, you can use the search tool with a query parameter and an optional search_recency_filter parameter. The server will return the search results content and citations.

Key features of Perplexity Search MCP

  • Provides a search tool for AI assistants

  • Uses Perplexity's chat completions API

  • Supports sonar and sonar-pro models

  • Allows filtering search results by recency (month, week, day, hour)

Use cases of Perplexity Search MCP

  • Enabling AI assistants to answer questions that require real-time information

  • Providing context for AI agents to have more informed conversations

  • Automatically researching topics based on user input

  • Integrating web search functionality into AI-powered applications

FAQ from Perplexity Search MCP

What is an MCP server?

An MCP server enables communication tools and protocols between different software systems or applications.

What models does the server support?

Currently it uses Perplexity's chat completions API, so it utilizes the sonar and sonar-pro models.

What is a Perplexity API key?

Perplexity AI is a company that creates web search models and AI based question and answering models.

Is the recency parameter compulsory?

no, but if you pass it, it has to be one of those filters (month, week, day, hour)

What do you do with the citations data?

The citations arrays are the resources where the results were derived and should be passed to the user for verification purposes