# Perplexity AI MCP

> Perplexity AI MCP connects any AI agent to Perplexity's advanced search and chat models, guaranteeing web-grounded answers with source citations. Stop relying on general LLMs that hallucinate facts; use our tools to run live searches and deep conversations using Sonar and other specialized models directly from your favorite client.

## Overview
- **Category:** ai-frontier
- **Price:** Free
- **Tags:** web-search, ai-chat, citations, real-time-data, natural-language-processing, search-api

## Description

This MCP connects your AI agent to Perplexity's powerful search engine, letting you get web-grounded answers through natural conversation. Instead of relying on a general chat model that might make up facts, this setup forces the AI to cite its sources, giving you real-time context and verifiable data.

Using this MCP means your AI becomes a true research assistant. You can run basic searches using the search tool, or dive deep with specialized models like Sonar Pro for enhanced responses. The entire catalog is managed by Vinkius, so once you connect through any compatible client, all these advanced capabilities are available.

It’s built for accuracy. Whether you need a quick check of current market data or a detailed analysis with step-by-step reasoning, the models provide citations and source URLs right in the response. No more switching between Google and ChatGPT; your AI handles it all.

## Tools

### chat
Sends a chat message to various Perplexity models and gets an answer with web citations.

### chat_pro
Sends a chat message specifically to the Sonar Pro model for enhanced, highly detailed responses with citations.

### chat_with_reasoning
Sends a message to the Sonar Reasoning model, forcing it to return a detailed step-by-step chain of thought and citations.

### chat_with_reasoning_pro
Sends a complex message to the Sonar Reasoning Pro model for deep analysis, providing both reasoning steps and citations.

### get_usage
Checks your API usage statistics so you can monitor consumption limits.

### list_models
Lists all available Perplexity models, showing their IDs and specific capabilities for use in other tools.

### search
Performs a web search using the Perplexity Search API and returns snippets with source URLs and citations.

### sonar
Sends a message to the core Sonar model, which provides answers grounded in real-time web search results.

## Prompt Examples

**Prompt:** 
```
Search the web for 'latest advances in quantum computing 2025'.
```

**Response:** 
```
Found recent articles on quantum error correction breakthroughs, Google's Willow chip achieving below-threshold error rates, and IBM's 1000+ qubit processors. Key sources include Nature, MIT Technology Review and Science Daily.
```

**Prompt:** 
```
Ask Sonar: What is the current price of Bitcoin?
```

**Response:** 
```
Bitcoin is currently trading at $87,234 USD according to CoinMarketCap data from today. The price has increased 3.2% in the last 24 hours with a market cap of $1.73 trillion. Sources: CoinMarketCap, CoinGecko.
```

**Prompt:** 
```
Send a chat to sonar-pro asking 'Explain how transformers work in NLP' with return_related_questions enabled.
```

**Response:** 
```
Sonar-pro responded with a detailed explanation of transformer architecture including self-attention, multi-head attention, positional encoding and the encoder-decoder structure. Related questions returned include: 'What is the difference between BERT and GPT?', 'How does self-attention work?' and 'What are transformer limitations?'
```

## Capabilities

### Search the Live Web
Runs dedicated searches against the web to pull current articles, snippets, and links.

### Generate Cited Responses
Engages chat models (sonar/sonar-pro) to answer questions using real-time data found online, including citations for every claim.

### Deep Reasoning and Analysis
Uses specialized Sonar reasoning models to break down complex topics into detailed, step-by-step explanations.

### Discover Model Options
Lists all available Perplexity AI models so you know exactly which depth or type of model you're running.

## Use Cases

### Validating a Research Thesis
A PhD candidate needs to prove three competing theories about climate change. Instead of relying on general chat, they use the search tool and then feed the results into the sonar model. The resulting output provides citations from Nature and Science Daily for every claim, allowing them to build their argument with verifiable data.

### Checking Today's Market Data
A financial analyst needs the current stock price of a niche company. They run a targeted search query using the domain filter set to 'financial-news.com'. The agent pulls up real-time market data and source links, which they can immediately incorporate into their report.

### Debugging Complex Code Logic
A developer is stuck on an obscure API integration error. They use the chat_with_reasoning model, asking it to explain the protocol failure step-by-step and citing documentation pages from known tech sources.

### Preparing a Quarterly Business Review
A marketing manager needs to summarize competitor movements. They run multiple focused searches (e.g., 'competitor X product launch 2025') using the search tool, gathering snippets and links from several sources in one pass.

## Benefits

- Accuracy is built in. When you use the chat tool, every factual claim comes paired with a citation and source URL, eliminating hallucination.
- Go beyond simple questions. The chat_with_reasoning model forces your agent to show its work, providing step-by-step reasoning for complex topics.
- Fine-tune your search results using the search tool's domain filter, letting you limit outcomes to specific types of websites (e.g., only academic journals).
- Access specialized depth with chat_pro and chat_with_reasoning_pro; these modes give you higher performance models for complex data extraction.
- Monitor your usage via the get_usage tool. You always know where you stand so you don't hit unexpected API limits mid-project.

## How It Works

The bottom line is that you get reliable, sourced information without having to leave your AI workflow.

1. Subscribe to this MCP and input your unique Perplexity API Key.
2. Connect the key to any compatible client (like Cursor or Claude).
3. Run a search query, and the agent will return web-grounded answers complete with source citations.

## Frequently Asked Questions

**How does the Perplexity AI MCP handle real-time data?**
The MCP uses dedicated search tools (like 'search' or 'sonar') that execute live queries against the web, ensuring answers are based on current information rather than old training data.

**Is the chat tool suitable for complex research? **
Yes. For deeper analysis, use chat_with_reasoning or chat_with_reasoning_pro. These models force the AI to show its reasoning steps before giving the final answer.

**What is the difference between 'sonar' and general chatting?**
General chat relies on trained knowledge; sonar uses a specialized process that forces web-grounding. It guarantees citations, which you won't get from basic conversation tools.

**How can I restrict the search results?**
The MCP allows you to use the search tool with domain filtering. You can limit your results to specific types of sites (e.g., only academic domains) for better accuracy.

**Do I need a separate API key for each model?**
No, you connect one API Key through the MCP, and it gives you access to all available models listed via list_models. You just select which tool or mode you want to use.