The Perplexity MCP server — perplexity-ask, maintained by Perplexity — lets an AI model hand a question to Perplexity's Sonar API and get back a real-time, web-grounded answer with citations. Instead of raw search links, the model receives a synthesised response it can quote and attribute, making it ideal for research and fact-checking agents.
Bring your own
Claude Sonnet 4.5
MCP Playground runs 40+ models on the same workflow: switch anytime, or use Compare mode to run several in parallel and balance quality vs. cost.
A Perplexity API key from perplexity.ai/settings/api. Deploy a sandboxed hosted instance from the MCP Playground catalog and paste the key at deploy time, or run the server yourself.
How models use it and what it is built for.
The Perplexity MCP server exposes the Sonar API as an MCP tool. When a model calls it, Perplexity runs a live web search, reads the top sources and returns a written answer with inline citations — all in one round trip. It runs as a Node.js stdio process configured with a Perplexity API key. Because the answer is already grounded and cited, it complements raw-result search servers like Brave: the agent can delegate "find the answer and cite it" to Perplexity, then verify or expand with other tools. The reference implementation is open source in Perplexity's modelcontextprotocol repository.
Typical tools an AI model can call. Exact names vary by version.
Copy any of these into MCP Agent Studio after connecting.
Ask Perplexity what changed in the EU AI Act enforcement timeline for 2026 and keep the citations.
What is the current recommended way to deploy a remote MCP server? Give me a cited summary.
Fact-check this claim with sources: "MCP was first released in late 2024."
Summarise the top three findings from recent research on retrieval-augmented generation, with links.
This is not a single-model product: you get the same MCP connection with 40+ models (Claude, GPT, Gemini, DeepSeek, open-weight, and more), you can switch mid-conversation, and you can open Compare mode to run the same prompt against multiple models at once. The card above is a suggested starting point for this server — not the only choice.
Default pick for Perplexity
Claude Sonnet 4.5
Claude Sonnet 4.5 knows when to delegate to Perplexity and how to weave the cited answer into a larger response. Haiku 4.5 works for simple single-question lookups.
Open MCP Agent Studio with the connection pre-filled. Add your token, pick any of 40+ models, and start chatting — no install required.
Open Agent StudioCommon questions about connecting, scoping and using it safely.
It is the official perplexity-ask MCP server from Perplexity. It lets any MCP-compatible AI model call Perplexity's Sonar API and receive a real-time, web-grounded answer with citations, rather than a list of raw search links.
The core tool is perplexity_ask, which forwards a conversation to the Sonar API and returns a synthesised answer with inline source citations. Some builds add reasoning and research variants, but perplexity_ask is the one every version ships.
A search server like Brave returns links and snippets and leaves the reading to your model. Perplexity does the searching, reading and synthesis on its side and hands back a finished, cited answer — useful when you want a grounded response in a single tool call.
You need a Perplexity API key with Sonar API access, which is billed per usage separately from a consumer Pro subscription. Create the key in your Perplexity API settings and paste it when you deploy the server.
Open MCP Agent Studio, deploy the hosted Perplexity server from the catalog, paste your Sonar API key, choose a model and ask a question. You can watch the perplexity_ask call and its cited answer stream back live.
Brave Search
Give an AI agent live web and local search via the Brave Search API.
Tavily
Search, extract and crawl the web for AI research agents via Tavily.
Wikipedia
Ground AI answers in verifiable Wikipedia articles — no API key needed.
Hacker News
Search and read Hacker News stories, comments and users from an AI agent.
