Perplexity's raw web search API: ranked results with titles, URLs, snippets and dates from Perplexity's own index, for grounding AI agents and RAG pipelines.
The Perplexity Search API, launched in September 2025, gives developers the raw search layer behind Perplexity: you send a query and get back ranked web results (title, URL, snippet, publication and last-updated dates) instead of a written answer. Perplexity says the results come from its own index of hundreds of billions of webpages, updated continuously. You decide what to do with the results, which makes it a building block for agents and retrieval-augmented generation (RAG) rather than a chatbot.
One request can carry up to five queries and still bills as a single request. Results can be filtered by domain (up to 20), language (up to 20 language codes), country, publication date range, last-updated date range and recency, and you can ask for between 1 and 20 results. Official SDKs exist for Python (perplexityai) and TypeScript (@perplexity-ai/perplexity_ai), and Perplexity lists integrations with n8n, Haystack, Mastra and Composio.
This is not the Sonar API. Perplexity ended Sonar Chat Completions on September 27, 2026 and now points developers who want a cited, written answer to its Agent API, which can call web search as a tool.
Perplexity Search API pricing
Option
Price
Notes
Search (standard) $5 per 1,000 requests
Up to 5 queries per request bill as one request; extracted text is not charged extra
Search (search_type: fast) $1 per 1,000 requests
Lower-latency mode
Agent API web_search tool
$2.50 per 1,000 calls
Plus model token costs, when search runs inside the Agent API
The pricing page does not state a free tier for the Search API. At 10,000 searches a month, standard search costs $50 and fast search $10.
Prices checked on the vendor's pricing page on November 10, 2026. See Perplexity API pricing.
Key features of Perplexity Search API
What Perplexity's own docs list:
Ranked web results with title, URL, snippet, date and last-updated date
Results from Perplexity's own index, not a resold Google or Bing feed
Multi-query: up to 5 queries in one request, billed once
Filters: up to 20 domains, up to 20 languages, country
Date filters: published after or before, last updated after or before, recency
1 to 20 results per query; fast mode for lower latency
SDKs for Python and TypeScript; integrations with n8n, Haystack, Mastra and Composio
Who the Perplexity Search API is best for
It suits developers who want raw, fresh results from an independent index with precise control over language, country and dates, for example a multilingual research agent or a news-monitoring pipeline. Fast mode at $1 per 1,000 requests is among the cheapest raw search on the market.
If you want meaning-based discovery of similar pages, Exa is the stronger fit. If you want a drop-in tool for LangChain-style agent frameworks, Tavily needs less glue. If you need Google's exact results pages, use a SERP API such as SearchApi or SerpApi. Our web search API comparison puts all eight side by side. Pros
+Fast mode costs $1 per 1,000 requests
+Own index, not a resold Google or Bing feed
+Richest language, country and date filters among search APIs
+Five queries per request for the price of one
Cons
−No free tier listed on the pricing page
−Returns results only; you build the answer layer yourself
−No published latency figures
Perplexity Search API FAQ
How much does the Perplexity Search API cost?
$5 per 1,000 requests for standard search and $1 per 1,000 with the fast search type. Up to five queries in one request bill as a single request. Prices checked on the vendor's pricing page on November 10, 2026.
Is the Perplexity Search API free?
Perplexity's pricing page does not list a free tier for the Search API. You pay per request from the first call.
What is the difference between the Perplexity Search API and Sonar?
The Search API returns raw ranked results that you process yourself. Sonar returned a written, cited answer; Perplexity ended Sonar Chat Completions on September 27, 2026 and now points that use case to its Agent API.
Does the Perplexity Search API use Google results?
No. Perplexity says results come from its own index of hundreds of billions of webpages, not from Google or Bing.
Which filters does the Perplexity Search API support?
Domain (up to 20), language (up to 20 codes), country, published-date range, last-updated range and recency, plus 1 to 20 results per query.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
Stork Quadrant
Becomes the API· 32/100
Replaceable as a UI, but kept alive as the API the agents call.
“This is a search index API competing against Google, Bing, Brave, and Exa — all of which have larger indexes, longer track records, and stronger developer trust. Perplexity doesn't own proprietary data; they crawl the same web. The only differentiator is index freshness and ranking quality, neither of which is a durable moat. This will get squeezed from above by Google and from below by cheaper alternatives.”
Defensibility · 0/100
Physical-world coupling
Regulatory moat
Network liquidity
Proprietary refreshing data
High-trust catastrophic workflows
Multi-party coordination
Brand / community / taste
An LLM alone could replace
Return ranked web results with titles, URLs, and snippets — Google Search API, Bing Search API, and Brave Search API all do this today Filter results by domain, date, and language — standard capability in every major search API
Ground an LLM response with fresh web content — any search API plus an LLM prompt does this
Power a RAG pipeline with web retrieval — commodity pattern, no Perplexity-specific advantage
Agent-Readiness · 70/100
·Verified MCP
✓Listed on agent surfaces— anthropic_directory, cursor
Stop selling raw search results and become the answer layer with verifiable citations that agents can trust — own the liability for factual grounding, not just the retrieval. Alternatively, build vertical indexes (legal, medical, financial) with proprietary crawl agreements that competitors can't replicate.
→Ship an MCP server and list it on Stork — biggest single point gain (+25).