
Perplexity
OfficialOfficial Perplexity MCP server — web search, ask, deep research, and reasoning via Sonar models.
Add to your client
Copy the config for your MCP client and paste it into its config file.
npx -y @perplexity-ai/mcp-serverPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"perplexity": {
"command": "npx",
"args": [
"-y",
"@perplexity-ai/mcp-server"
],
"env": {
"PERPLEXITY_API_KEY": "<your-perplexity-api-key>"
}
}
}
}Before you start
- Node.js with npx available
- A Perplexity API key from the API portal at https://www.perplexity.ai/account/api/group (set as PERPLEXITY_API_KEY)
- API access requires adding a payment method / credits to your Perplexity account
About Perplexity
This is Perplexity's official MCP server, giving AI assistants real-time web search, conversational answers, deep research, and structured reasoning powered by Perplexity's Sonar models. It is published on npm as @perplexity-ai/mcp-server and runs over stdio (with an HTTP server mode available for cloud deployments).
It exposes four tools mapped to different Sonar capabilities: perplexity_search for ranked web search results, perplexity_ask for conversational answers with live web search (sonar-pro), perplexity_research for comprehensive reports (sonar-deep-research), and perplexity_reason for complex analytical tasks (sonar-reasoning-pro).
The research and reasoning tools accept an optional strip_thinking parameter to drop intermediate reasoning markup and save context tokens. The server also supports proxy and timeout configuration via environment variables (PERPLEXITY_PROXY, PERPLEXITY_TIMEOUT_MS), which is handy in corporate environments and for long-running deep-research calls.
Tools & capabilities (4)
perplexity_searchDirect web search via the Perplexity Search API, returning ranked search results.
perplexity_askGeneral-purpose conversational answers with real-time web search using the sonar-pro model.
perplexity_researchDeep, comprehensive research and detailed reports using the sonar-deep-research model.
perplexity_reasonAdvanced reasoning and problem-solving for complex analytical tasks using sonar-reasoning-pro.
When to use it
- Use it when you want grounded, up-to-date answers with citations instead of the model's stale training data.
- Use it when you need a quick ranked list of web results to feed into an agent (perplexity_search).
- Use it when you need a long, sourced research report on a topic (perplexity_research / sonar-deep-research).
- Use it when a task needs structured multi-step reasoning rather than a simple lookup (perplexity_reason).
- Use it when you want to strip chain-of-thought tags from responses to save context tokens (strip_thinking).
Quick setup
- 1Create a Perplexity API key at https://www.perplexity.ai/account/api/group.
- 2Set the PERPLEXITY_API_KEY environment variable with your key.
- 3Add the server to your client config with command `npx -y @perplexity-ai/mcp-server`, or run `claude mcp add perplexity --env PERPLEXITY_API_KEY=... -- npx -y @perplexity-ai/mcp-server`.
- 4Optionally configure PERPLEXITY_TIMEOUT_MS and PERPLEXITY_PROXY for long research calls or corporate networks.
- 5Restart your MCP client and verify by asking it to run a Perplexity search.
Security notes
Requires a PERPLEXITY_API_KEY that incurs usage charges; keep it secret and rotate if exposed. Returned answers summarize live web content, so verify citations before relying on them for high-stakes decisions.
Perplexity FAQ
Is the Perplexity API free?
No. The Sonar API is usage-based and requires a payment method/credits on your Perplexity account; there is no fully free tier for the MCP server's API calls.
What is the difference between perplexity_research and perplexity_reason?
Research uses sonar-deep-research for thorough, report-style analysis, while reason uses sonar-reasoning-pro for structured multi-step problem solving on complex analytical tasks.
What does the strip_thinking parameter do?
On the research and reasoning tools, setting strip_thinking: true removes intermediate thinking tags from the response, reducing token usage in your client's context.
My deep research requests time out — how do I fix it?
Increase PERPLEXITY_TIMEOUT_MS (the default is around five minutes); deep-research calls can be slow, and the timeout is configurable via that env var.
Which clients does it work with?
Any MCP-compatible client over stdio, including Claude Desktop, Cursor, and VS Code; an HTTP server mode is also available for cloud deployments.
Alternatives to Perplexity
Official MCP reference server that fetches a URL and returns its content as clean Markdown, with chunking.
Official Firecrawl MCP server — scrape, crawl, map, search, and structured extraction for any LLM client.
Exa's neural web search and crawling MCP server — runs locally via npx or as a hosted remote endpoint.
Compare Perplexity with: