@perplexity-ai/mcp-server

What it does

This is the official MCP server for Perplexity AI's API platform. It connects Claude to Perplexity's real-time search and reasoning models via four tools: direct web search, conversational search backed by Sonar Pro, deep research using Sonar Deep Research, and advanced reasoning with Sonar Reasoning Pro. All operations delegate to Perplexity's API, which handles the actual web search and model inference.

Who it's for

Researchers and engineers building AI agents or chatbots that need grounding in current information and source citations. Also useful for anyone who wants Claude to answer questions about recent events, products, or data without hallucinating based on stale training data.

Common use cases

• Query current news, product launches, or market data that post-dates the model's knowledge
• Generate detailed research reports with citations from live sources
• Solve complex analytical or reasoning problems grounded in real-time context
• Build fact-checking tools that verify claims against current information
• Create conversational assistants that cite sources and stay up-to-date

Setup pitfalls

• Requires a valid Perplexity API key from api.perplexity.ai; all requests fail without valid credentials
• Synchronous HTTP calls have a default 5-minute timeout; long-running research queries may need PERPLEXITY_TIMEOUT_MS increased
• If behind a corporate proxy, set PERPLEXITY_PROXY explicitly; default proxy detection may not work reliably
• The strip_thinking parameter for reducing context token usage is only available on the reasoning and research tools, not search or ask