Arjunkmrm Perplexity Search

What arjunkmrm-perplexity-search MCP server does

The arjunkmrm-perplexity-search MCP server exposes Perplexity web search to AI assistants through the Model Context Protocol. It uses Perplexity's chat completions API and supports the sonar and sonar-pro models.

The main operation is a search request. An assistant provides a query, and the server returns search-result content together with a citations array. This makes the server suitable for research questions where the assistant needs web information and source references in the response.

How it works

The arjunkmrm-perplexity-search MCP server receives the search request from an MCP client, passes the request to Perplexity, and returns the resulting content and citations. A query is required for every search.

Searches can optionally include a recency filter. Supported values are month, week, day, and hour. If no filter is supplied, the search does not apply a time restriction. The available MCP inputs are represented by query and search_recency_filter; the README describes these as parameters of the search tool.

Setup and configuration

Set the PERPLEXITY_API_KEY environment variable before running the server. The key is required because requests are sent to Perplexity's API.

No additional environment variables, model settings, or client-specific configuration are documented in the provided material. After the API key is available, add the server to an MCP-compatible assistant using that client's normal server configuration process.

Tools and capabilities

The arjunkmrm-perplexity-search MCP server provides:

• A search tool for web searches through Perplexity.
• A required query input containing the search request.
• An optional search_recency_filter input with month, week, day, or hour options.
• Support for Perplexity's sonar and sonar-pro models.
• Responses containing search content and citations.

The material does not document separate operations beyond the search capability or describe controls for selecting between the two models.

Limitations and notes

This server depends on a valid Perplexity API key and the Perplexity chat completions API. Without the required key, it cannot make searches. Recency filtering is limited to the four documented periods, and omitting the filter leaves the search unrestricted by time.

The provided documentation does not identify specific MCP clients, transport details, rate limits, result-count controls, or behavior for API errors. Those details should be verified in the repository or the deployment environment before relying on them in production. The project is licensed under MIT.