Perplexity

Perplexity is an MCP server that provides Model Context Protocol clients with direct access to Perplexity's specialized search, reasoning, and research AI models. Built for developers, technical analysts, and researchers using tools like Claude or Cursor, it expands agent capabilities with live web search and deep online information gathering. The server integrates three distinct tools mapped to specific Perplexity models: quick search via Sonar Pro for straightforward queries, multi-step analysis via Sonar Reasoning Pro for complex explanations and comparisons, and Sonar Deep Research for exhaustive reports across specific focus areas. It incorporates automatic query complexity detection to route prompts to the appropriate model, while also supporting explicit model overrides. By connecting an agent to Perplexity APIs, users can conduct targeted external research, verify facts, analyze engineering trade-offs, and generate comprehensive research summaries directly inside their development environments and MCP clients without manual web browsing.

Category: AI & LLM Tooling

Tags: perplexity, reasoning, research, search

Visit Perplexity

How to install and configure Perplexity

1. Obtain a Perplexity API key from your Perplexity account settings. 2. Ensure Node.js is installed on your system. 3. Add the server configuration to your MCP settings file. For NPX execution without local cloning: json { "mcpServers": { "perplexity": { "command": "npx", "args": ["-y", "perplexity-mcp"], "env": { "PERPLEXITY_API_KEY": "your_api_key" } } } } 4. Alternatively, clone the repository, build the project, and run it using node /path/to/perplexity-server/build/index.js. 5. If your client cannot read environment variables, pass the key via the --api-key your_key command-line argument or specify a .env directory using --cwd /path/to/project.

What you can do with Perplexity

• Conducting quick factual lookups and retrieving concise answers using the Sonar Pro search tool. * Analyzing architectural trade-offs, technical comparisons, and step-by-step problem-solving using the Sonar Reasoning Pro tool. * Generating multi-source, in-depth research reports with customized focus areas using the Sonar Deep Research tool. * Routing user queries automatically to the appropriate model based on query complexity detection or via forced overrides.

Key facts

• https://github.com/DaInfernalCoder/perplexity-mcp
• AI & LLM Tooling, Web Search & Research
• perplexity, reasoning, research, search

Part of MCP Servers

What tools are included in Perplexity MCP Server?

The server includes three primary tools: search, reason, and deep_research. The search tool uses Sonar Pro for fast information lookup on simple queries. The reason tool uses Sonar Reasoning Pro for complex, multi-step evaluations and technical comparisons. The deep_research tool uses Sonar Deep Research to perform comprehensive investigations across specified focus areas.

How do I install Perplexity MCP Server?

You can configure it via NPX by specifying the npx command, the perplexity-mcp package, and setting your PERPLEXITY_API_KEY environment variable in your client settings. Alternatively, you can clone the repository from GitHub, compile the files, and run the server script directly using Node.js.

How can I provide the Perplexity API key if environment variables fail?

The server provides fallback configuration methods if your MCP client cannot parse environment variables. You can pass the API key directly using the --api-key command-line argument, or point the server to a directory containing a valid .env file using the --cwd argument. Command-line flags take priority over environment variables.

Can I force a specific Perplexity model for a query?

Yes. While the server includes automatic query complexity detection to select the best model, each tool supports an optional force_model argument. Setting this boolean parameter to true forces the MCP server to execute the prompt using that specific tool regardless of the query length or complexity.