Perplexity MCP Server icon

Perplexity MCP Server

Official

The official MCP server implementation for Perplexity API Platform. Provides real-time web search, deep research, and advanced reasoning capabilities through Sonar AI models and the Search API.

OfficialSearchAI IntegrationResearch
View on GitHub

Perplexity MCP Server

The Perplexity MCP Server is the official Model Context Protocol (MCP) implementation for the Perplexity API Platform. It provides AI assistants with real-time web search, deep research, and advanced reasoning capabilities through Perplexity's powerful Sonar AI models and Search API.

Why Use Perplexity MCP Server?

Perplexity MCP Server bridges the gap between AI assistants and the live web, enabling:

  • Real-time information access: Search the web for current events, news, and up-to-date documentation
  • Grounded responses: Get answers backed by actual web sources with citations
  • Deep research capabilities: Conduct thorough investigations with comprehensive analysis
  • Advanced reasoning: Solve complex problems with step-by-step logical analysis

Available Tools

The server exposes four powerful tools backed by Perplexity's Sonar AI models:

1. perplexity_search

Direct web search using the Perplexity Search API. Returns ranked search results with titles, URLs, snippets, and metadata.

Best for: Finding current information, news, facts, documentation, or specific web content.

2. perplexity_ask

General-purpose conversational AI with real-time web search using the sonar-pro model (200k context window).

Best for: Quick questions, everyday searches, and conversational queries requiring current information.

3. perplexity_research

Deep, comprehensive research using the sonar-deep-research model. Provides thorough analysis with full citations.

Best for: Complex topics requiring detailed investigation, academic research, and in-depth analysis.

4. perplexity_reason

Advanced reasoning and problem-solving using the sonar-reasoning-pro model with chain-of-thought processing.

Best for: Logical problems, complex analysis, decision-making, and tasks requiring step-by-step reasoning.

Tip: Use the strip_thinking parameter to remove <think>...</think> tags and conserve context tokens.

Installation

One-Click Installation

Perplexity MCP Server offers one-click installers for popular MCP clients:

  • Cursor: Automatic configuration available
  • VS Code: Direct installation support
  • Claude Desktop: JSON configuration import

Claude Code Installation (Recommended)

# Set your API key
export PERPLEXITY_API_KEY="your_key_here"

# Launch Claude Code
claude

# Install via plugin marketplace
/plugin marketplace add perplexityai/modelcontextprotocol
/plugin install perplexity

NPX Installation

For any MCP-compatible client, add this to your configuration:

{
  "mcpServers": {
    "perplexity": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "your_key_here"
      }
    }
  }
}

Docker Deployment

For production or isolated environments:

docker build -t perplexity-mcp-server .
docker run -p 8080:8080 -e PERPLEXITY_API_KEY=your_key_here perplexity-mcp-server

Server endpoint: http://localhost:8080/mcp

Configuration

Required Environment Variables

| Variable | Description | Required | |----------|-------------|----------| | PERPLEXITY_API_KEY | Your Perplexity API key | Yes |

Optional Environment Variables

| Variable | Description | Default | |----------|-------------|---------| | PERPLEXITY_TIMEOUT_MS | Request timeout in milliseconds | 600000 (10 min) | | PERPLEXITY_PROXY | Proxy URL for corporate networks | None | | PORT | HTTP server port | 8080 | | BIND_ADDRESS | Server bind address | 127.0.0.1 | | ALLOWED_ORIGINS | CORS allowed origins | None |

Getting Your API Key

  1. Visit the Perplexity API Portal
  2. Sign in or create an account
  3. Navigate to API Keys section
  4. Generate a new API key
  5. Copy and store securely

Use Cases

Research and Documentation

Search for the latest React 19 features and migration guide

Real-Time Information

What are the current cryptocurrency prices for Bitcoin and Ethereum?

Technical Problem-Solving

Analyze the trade-offs between using MongoDB vs PostgreSQL for a social media application with 1 million users

Code Research

Research best practices for implementing WebSocket connections in Next.js 15

Technical Details

Best Practices

Choosing the Right Tool

Use perplexity_search when:

  • You need current information (news, stock prices, weather)
  • You want ranked search results with URLs
  • You need to verify facts with sources
  • You're looking for specific documentation or resources

Use perplexity_ask when:

  • You need quick answers to everyday questions
  • You want conversational responses with context
  • The query benefits from AI interpretation
  • You need general information synthesis

Use perplexity_research when:

  • Conducting comprehensive analysis of complex topics
  • You need detailed citations and sources
  • Research requires thorough investigation
  • Building technical documentation or reports

Use perplexity_reason when:

  • Making architectural or technical decisions
  • Analyzing trade-offs between solutions
  • Solving complex logical problems
  • Need step-by-step problem-solving

Optimization Tips

  1. Context Tokens: Use strip_thinking: true with perplexity_reason to reduce token usage
  2. Timeout Settings: Increase PERPLEXITY_TIMEOUT_MS to 900000 (15 min) for deep research queries
  3. Rate Limiting: Implement retry logic for API rate limits
  4. Caching: Cache frequent queries locally to reduce API costs
  5. Query Refinement: Be specific in prompts for better results

Integration Patterns

For Development Workflows:

"Search for the latest security best practices for React authentication in 2025"

For Research Tasks:

"Research the performance implications of using WebSockets vs Server-Sent Events for real-time data"

For Code Documentation:

"Find current documentation for TypeScript 5.5 utility types and provide examples"

For Decision Making:

"Analyze whether I should use PostgreSQL or MongoDB for a multi-tenant SaaS application"

Integration with MCP Clients

Roo Code Integration

Perplexity MCP Server works seamlessly with Roo Code's custom modes:

  1. Create a Research Mode in Roo Code settings
  2. Assign Perplexity to the mode's available tools
  3. Use @perplexity in conversations to trigger research
  4. Combine with code editing for research-driven development

Claude Desktop Integration

{
  "mcpServers": {
    "perplexity": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "pplx-xxxxxxxxxxxxxxxxxxxxxxxxxxxx"
      }
    }
  }
}

Save to claude_desktop_config.json and restart Claude Desktop.

VS Code with Cline/Continue

Add to your MCP configuration:

{
  "mcpServers": {
    "perplexity": {
      "command": "npx",
      "args": ["-y", "perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "your_api_key_here"
      }
    }
  }
}

Performance Considerations

Response Times

  • perplexity_search: 1-3 seconds (fast)
  • perplexity_ask: 3-8 seconds (moderate)
  • perplexity_research: 2-5 minutes (comprehensive)
  • perplexity_reason: 10-30 seconds (with thinking), 5-10 seconds (stripped)

Cost Optimization

  1. Use search first for simple lookups (lower cost)
  2. Reserve deep research for complex topics (higher cost)
  3. Cache results when possible to avoid duplicate queries
  4. Use reasoning selectively for important decisions
  5. Monitor API usage through Perplexity dashboard

API Limits

  • Free Tier: Limited requests per month
  • Standard: 5,000 requests/month
  • Pro: 20,000+ requests/month
  • Enterprise: Custom limits

Check current limits at Perplexity API Portal.

Comparison with Other Search Tools

| Feature | Perplexity MCP | Google Search API | Bing Search API | |---------|----------------|-------------------|-----------------| | AI-Powered | ✅ Yes | ❌ No | ⚠️ Limited | | Real-Time Web | ✅ Yes | ✅ Yes | ✅ Yes | | Deep Research | ✅ Yes | ❌ No | ❌ No | | Reasoning | ✅ Advanced | ❌ No | ❌ No | | Citations | ✅ Full | ⚠️ Links only | ⚠️ Links only | | MCP Native | ✅ Yes | ❌ No | ❌ No | | Cost | $$ | $$ | $$ |

Frequently Asked Questions

Can I use Perplexity MCP Server without a paid API key?

No, you need a Perplexity API key which requires a paid subscription. Visit https://www.perplexity.ai/account/api to sign up.

Does it work with local LLMs?

Yes! The MCP server acts as a tool that any MCP-compatible client can use, regardless of which LLM powers the client (OpenAI, Anthropic, local models like Ollama, etc.).

What's the difference between perplexity_ask and perplexity_search?

  • perplexity_search: Returns raw search results (titles, URLs, snippets) - you get the data to analyze yourself
  • perplexity_ask: Uses AI to synthesize an answer from search results - you get a conversational response with citations

Can I use it in production applications?

Yes, the Perplexity MCP Server can be deployed in production. Use Docker for containerization and set BIND_ADDRESS=0.0.0.0 for public hosting. Ensure proper API key management and rate limiting.

How do I reduce costs?

  1. Use perplexity_search for simple lookups (cheaper)
  2. Enable strip_thinking: true for reasoning queries
  3. Cache frequent queries
  4. Use appropriate timeouts to avoid long-running queries
  5. Monitor usage in Perplexity dashboard

Is my search data private?

Search queries are sent to Perplexity's API. Review Perplexity's Privacy Policy for details on data handling.

Can I self-host the MCP server?

Yes! The server is open-source. Clone the GitHub repository, build from source, and run locally. You still need a Perplexity API key for the backend.

What models power the different tools?

  • perplexity_search: Direct Search API (no model)
  • perplexity_ask: sonar-pro (200k context)
  • perplexity_research: sonar-deep-research
  • perplexity_reason: sonar-reasoning-pro

Advanced Configuration

Corporate Proxy Support

For enterprise environments:

export PERPLEXITY_PROXY="https://username:[email protected]:8080"
export PERPLEXITY_TIMEOUT_MS=900000

Custom Port for HTTP Deployment

docker run -p 3000:3000 \
  -e PORT=3000 \
  -e PERPLEXITY_API_KEY=your_key \
  -e BIND_ADDRESS=0.0.0.0 \
  perplexity-mcp-server

Multiple Instances

Run multiple server instances for different purposes:

{
  "mcpServers": {
    "perplexity-research": {
      "command": "npx",
      "args": ["-y", "perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "key_for_research",
        "PERPLEXITY_TIMEOUT_MS": "900000"
      }
    },
    "perplexity-quick": {
      "command": "npx",
      "args": ["-y", "perplexity-mcp"],
      "env": {
        "PERPLEXITY_API_KEY": "key_for_quick_search",
        "PERPLEXITY_TIMEOUT_MS": "30000"
      }
    }
  }
}

Real-World Use Cases

Software Development

Staying Current with Technology:

"Search for breaking changes in Next.js 15 that affect App Router"

API Research:

"Research the latest Stripe API features for subscription management released in 2025"

Debugging Assistance:

"Search for solutions to CORS errors in Express.js with credentials"

Technical Writing

Fact Verification:

"Verify the current system requirements for running Docker Desktop on Windows 11"

Industry Trends:

"Research the adoption rate of TypeScript vs JavaScript in 2025"

Product Management

Competitive Analysis:

"Research the latest features announced by Vercel and Netlify for edge computing"

Market Research:

"Analyze current trends in AI coding assistants and developer tools"

DevOps & Infrastructure

Best Practices:

"Research current best practices for Kubernetes security in production"

Tool Comparison:

"Compare Terraform vs Pulumi for infrastructure as code in 2025"

Resources

Use Perplexity MCP Server MCP Server with

5ire

Open source cross-platform desktop AI assistant with native MCP server support for enhanced tool integration.

desktopopen-source+1

Claude.ai

Anthropic's official web client with MCP server integration for enhanced AI assistant capabilities and tool connectivity.

anthropicweb-client+1

Claude Desktop App

Official Claude desktop application with built-in MCP support and comprehensive server management capabilities

anthropicdesktop-app+1

Cline

Advanced agentic coding VS Code extension that leverages MCP servers for powerful development tools and automation

vscodecoding-assistant+1

Continue

Open-source development copilot for VS Code and JetBrains IDEs with comprehensive MCP integrations and extensible architecture

open-sourceide-extension+1

Cursor

Advanced AI-powered IDE with comprehensive MCP support for tools and resources, designed for modern development workflows

ideai-powered+1

Daydreams Agents

Innovative agentic platform with Model Context Protocol support for building and deploying autonomous AI agents

agentsplatform+1

Emacs MCP

Community-driven Emacs integration packages for Model Context Protocol support and tool integration

emacseditor+1

fast-agent

Ultra-lightweight agent framework with Model Context Protocol support designed for high-performance AI agent deployment

agent-frameworklightweight+1

FLUJO

Open-source MCP Hub and Inspector with multi-model workflow orchestration and intuitive chat interface

open-sourceworkflow+1

AgentAI

Rust library for simplified AI agent creation with built-in MCP integration and high-performance async capabilities.

rustlibrary+1

Genkit

Google's powerful Genkit framework with comprehensive MCP support for building AI-powered applications and workflows

googleframework+1

Glama

Comprehensive MCP directory, client listings, and community hub for discovering and sharing MCP resources

directorycommunity+1

GenAIScript

Microsoft's powerful scripting framework with comprehensive MCP client integration for AI-powered automation and scripting

microsoftscripting+1

Goose

Open-source development agent with comprehensive MCP integration and intelligent IDE automation capabilities

open-sourcedev-agent+1

gptme

Versatile terminal and chat client with comprehensive MCP support for both local and hosted AI models

terminalchat-client+1

HyperAgent

Playwright-powered autonomous web agent with comprehensive MCP tools integration for automated web interactions

web-automationplaywright+1

JetBrains AI Assistant

JetBrains AI Assistant with full MCP client and server support. Connect to external MCP servers for enhanced AI capabilities, or expose 25+ IDE tools to external clients like Claude Desktop and Cursor.

IDEAI Integration+2

Klavis AI

Open-source MCP integration platform that connects AI agents to 100+ tools (GitHub, Slack, Notion, etc.) with enterprise-grade OAuth and intelligent context management. YC-backed, SOC 2 compliant.

platformintegration+2

LibreChat

Self-hosted, open-source ChatGPT-style UI with comprehensive MCP support, multi-model AI integration (OpenAI, Anthropic, Google, AWS), agents, code interpreter, and enterprise features. Fully customizable and privacy-focused.

web-appself-hosted+2

Lutra AI

First MCP-compatible client built for everyone. AI task agent that connects all your work apps, turns conversations into automated workflows, and integrates with MCP servers seamlessly. No-code automation platform.

workflow-automationno-code+2

AgenticFlow

No-code platform for building intelligent agents that connect thousands of APIs and tools via Model Context Protocol integration

no-codeagent-builder+1

mcp-agent

Simple, composable Python framework to build production-ready AI agents using MCP. Pairs Anthropic Building Effective Agents patterns with MCP runtime, supports multiple LLM providers, and includes Temporal-based orchestration for workflows.

frameworkpython+2

mcp-use

Open-source library to connect any LLM to any MCP server in 6 lines of code. Language-agnostic protocol supporting Python and TypeScript with LangChain.js compatibility, HTTP/SSE transport, and everything needed to build with MCP servers/clients and AI agents.

librarypython+3

MCPHub (mcp.nvim)

Neovim plugin to manage and use MCP servers within the editor. Features intuitive UI for managing server states, marketplace integration for exploring MCP servers, interactive testing, and integration with Avante.nvim, CodeCompanion.nvim, and CopilotChat.nvim.

editor-pluginneovim+2

MCPOmni-Connect

Versatile CLI client for MCP with agentic and orchestrator capabilities. Supports chat mode, autonomous agents, planner-based orchestration, treats each MCP server as tool agent, multi-agent workflows, and can be embedded in FastAPI or backend systems.

cliagent-framework+3

Memex

All-in-one desktop app and server builder with prompt-to-MCP server capabilities. Build custom MCP servers from natural language, integrate them instantly, and use them to solve problems—all within a single conversation.

desktop-appdeveloper-tools+2

MindPal

No-code platform for AI agents and workflows with comprehensive MCP support. Build custom AI agents trained on your data, create multi-agent workflows, and connect to 1,000+ business tools—all without writing code.

no-codeworkflow-automation+2

MooPoint

Web-based, model-agnostic AI chat platform with comprehensive MCP server integration, tool calling capabilities, and OAuth support. Connect to any LLM provider while maintaining control over your API keys and data.

web-appmulti-model+2

Msty Studio

Privacy-first AI workspace integrating local and online LLMs with comprehensive MCP toolbox. Run models entirely offline or use cloud providers while maintaining complete control over your data and workflows.

desktop-appprivacy-focused+2

NVIDIA Agent Intelligence Toolkit

Enterprise-grade toolkit for building AI agents with MCP client/server support. Works with LangChain, LlamaIndex, CrewAI, and Semantic Kernel. Includes observability, profiling, and hyperparameter tuning. Part of NVIDIA NeMo family.

platformframework+2

Amazon Q CLI

Amazon's AI-powered coding assistant for terminals with comprehensive MCP server support and AWS integration

awscli+1

OpenSumi

AI-native IDE framework for building custom development environments with MCP client support. Powers CodeFuse IDE and CodeBlitz. VSCode plugin compatible, customizable UI, supports containers, Electron, and pure frontend deployments. By Alibaba & Ant Group.

ideframework+2

oterm

A text-based terminal client for Ollama that supports the Model Context Protocol (MCP), allowing interaction with external tools and data sources directly from the CLI.

cliopen-source+2

Postman

The world's leading API platform now includes a native MCP client for testing, debugging, and interacting with MCP servers in a visual environment.

developer-toolstesting+2

Roo Code

Open-source AI-powered agentic coding assistant for VS Code with comprehensive MCP server support, multi-file editing, custom modes, and advanced orchestration features. Free to use with any AI model.

VS Code ExtensionAI Coding Assistant+2

Slack MCP Client

An open-source TypeScript client that bridges Slack with the Model Context Protocol, enabling users to interact with MCP tools directly within Slack channels and DMs.

chatintegration+2

Sourcegraph Cody

An advanced AI coding assistant that leverages the Model Context Protocol (via OpenCtx) to fetch context from external sources like issue trackers, databases, and documentation.

coding-assistantide-extension+2

SpinAI

A lightweight, observable TypeScript agent framework designed for building production-grade AI agents that can easily consume Model Context Protocol (MCP) tools.

frameworktypescript+2

Superinterface

A comprehensive developer platform and infrastructure for building in-app AI assistants that natively supports the Model Context Protocol (MCP) for tool integration.

platforminfrastructure+2

Superjoin

An AI Assistant for Google Sheets that functions as an MCP Client, allowing users to import data from external sources and automate spreadsheet workflows using natural language.

productivityspreadsheet+2

Theia AI

An open and flexible platform for building AI-enhanced tools and IDEs, featuring deep integration with the Model Context Protocol to provide rich context to AI agents.

platformframework+2

Apify MCP Tester

Standalone SSE/HTTP client for comprehensive MCP server testing with intuitive chat-like user interface

testingdebugging+1

Theia IDE

The open-source IDE that serves as the reference implementation for Theia AI, offering out-of-the-box support for the Model Context Protocol (MCP).

ideopen-source+2

Tome

A sleek macOS desktop client by Runebook that simplifies the management of local LLMs and Model Context Protocol (MCP) servers for non-technical users.

desktop-appmacos+2

TypingMind

A versatile web and desktop frontend for LLMs that supports the Model Context Protocol via a local bridge connector, allowing web-based agents to access local tools.

frontendweb-app+2

Warp

A modern, Rust-based terminal that integrates with Claude Code and the Model Context Protocol to bring AI agent capabilities directly into your command-line workflow.

terminalcli+2

WhatsMCP

Connect your personal AI stack and MCP servers directly to WhatsApp. Manage tools, agents, and workflows through a familiar chat interface.

messagingmobile+2

Windsurf Editor

The first agentic IDE, powered by Codeium. Features deep context awareness and native MCP support to connect to your external tools and data.

idedeveloper-tools+2

Witsy

A powerful desktop AI assistant for macOS and Windows. Supports local LLMs (Ollama) and cloud providers (OpenAI, Anthropic), with a native plugin system built on MCP.

desktopproductivity+3

Zed

A high-performance, multiplayer code editor from the creators of Atom. Features built-in MCP support to connect AI assistants to your codebase and external tools.

idedeveloper-tools+3

Zencoder

An AI coding agent and IDE that integrates with your entire software development lifecycle. Features a built-in MCP library for one-click tool installation.

idedeveloper-tools+3

Augment Code

Agentic code assistant that orchestrates MCP servers for repository-level tasks with intelligent automation

code-assistantautomation+1

Raycast MCP Extension

A Raycast extension that bridges the Model Context Protocol with Raycast AI. Access your MCP tools directly from the Raycast launcher and Quick AI.

launcherproductivity+3

LM Studio

A desktop application for discovering, downloading, and running local LLMs. Version 0.3.17+ includes native support for MCP, allowing local models to use tools.

desktoplocal-llm+3

Perplexity

Native macOS integration enabling Perplexity to connect with local apps and data (Notes, Calendar, Files) via the Model Context Protocol using the PerplexityXPC helper.

desktop-appsearch+2

ChatMCP

An open-source, feature-rich desktop chat client built with Flutter that implements the Model Context Protocol, featuring an integrated MCP server marketplace and multi-LLM support.

desktop-appopen-source+2

MCP Inspector

The official web-based debugging tool for Model Context Protocol servers, allowing developers to test tools, resources, and prompts interactively.

developer-toolstesting+3

mcphub.nvim

Neovim plugin to manage, test, and use MCP servers within the editor. Integrates seamlessly with AI chat plugins like Avante.nvim, CodeCompanion.nvim, and CopilotChat.nvim.

neovimide+2

Bolt.new

AI-powered full-stack development platform that creates stunning apps and websites through conversational AI, with native MCP server integration for extensible capabilities.

web-developmentai-powered+3

Rovo Dev CLI

An AI agent built by Atlassian for software developers, bringing AI-assisted development directly to the command line interface.

cliai-assistant+1

Toolprint

Build AI Agents that use tools 10x Better. A single SDK with powerful capabilities to teach and improve tool selection to achieve workflow goals.

ai-agentstool-management+2

Microsoft Copilot Studio

Low-code platform for building custom AI copilots and agents with MCP integration. Connect existing knowledge servers and data sources directly within Copilot Studio to extend agent capabilities.

enterpriselow-code+2

BeeAI Framework

Open-source framework with comprehensive MCP support for building sophisticated AI workflows and automations

frameworkai-workflows+1

GitHub Copilot (VS Code)

AI-powered code completion and chat in Visual Studio Code with native Model Context Protocol support. Access external tools and data sources through MCP to enhance code generation and assistance.

idecode-editor+2

Amp

Frontier coding agent for terminal and editor with multi-model support (Claude, GPT-5), native MCP integration, thread-based conversations, and extensions for VS Code, Cursor, JetBrains, and Neovim.

clieditor-extension+2

OpenCode

Open-source AI coding agent for the terminal with native MCP support, Plan/Build modes, undo/redo capabilities, and conversation sharing. Supports multiple LLM providers.

cliopen-source+2

BoltAI

Native macOS desktop AI application with comprehensive MCP integrations for both local and cloud-based AI models

macosdesktop-app+1

Claude Code

Anthropic's official AI coding assistant CLI and terminal application with comprehensive MCP support and development tools

anthropiccli+1

Related Servers

Vantage

Interact with your organization's cloud cost spend.

Official

VariFlight

VariFlight's official MCP server provides tools to query flight information, weather data, comfort metrics, the lowest available fares, and other civil aviation-related data.

Official

VCAgents

Interact with investor agents—think Wilson or Thiel—continuously updated with market intel.

Official

Vectorize

Vectorize MCP server for advanced retrieval, Private Deep Research, Anything-to-Markdown file extraction and text chunking.

Official

Verbwire

Deploy smart contracts, mint NFTs, manage IPFS storage, and more through the Verbwire API

Official

Vercel

Access logs, search docs, and manage projects and deployments.

Official
Sponsored
Algolia Inc logo

Algolia Inc

“I can't code” is no longer an excuse. The technical barrier is gone. What will you create first?

Try Algolia MCP

Related Categories

OfficialSearchAI IntegrationResearch