The Perplexity MCP Server is the official Model Context Protocol server for the Perplexity API platform, letting AI assistants and agent frameworks call Perplexity’s search-grounded models as tools. The GitHub repository, maintained by Perplexity AI, has collected around 2,580 stars and is actively updated in 2026. With it, any MCP-compatible client can ask questions that Perplexity answers with live web research and citations.
The server exposes Perplexity’s Sonar models through standard MCP tools, including general search-answering, a reasoning-focused variant, and a deep research mode that produces longer, report-style answers. Configuration is a standard MCP server entry with your Perplexity API key, after which clients such as Claude Desktop, Cursor, and other agent tools can ground their responses in current web information instead of relying only on training data. Official documentation covers installation for the major MCP hosts.
The server supports Perplexity’s Sonar model family, including the pro and reasoning variants, and its deep research tool can compile multi-source reports that would take a human analyst far longer to assemble. Setup guides cover the most common MCP hosts with copy-paste configuration, and environment-variable based key handling keeps credentials out of shared config files. For agent builders, it is one of the simplest ways to add trustworthy, up-to-date web grounding to an existing MCP setup.
The MCP server itself is free and open source under the MIT license. Usage is billed through your Perplexity API account under its normal per-request pricing.
Developers building agents and assistant workflows who want to give their models reliable, citation-backed access to current web knowledge through a standard protocol.