A concise guide to Perplexity AI search access, Search API results, Agent API citations and documented request limits, with practical suggestions for choosing the appropriate route.
What Perplexity AI search can mean
Perplexity uses the word “search” across several access routes and developer tools, so it helps to identify the intended output before choosing one. The official Perplexity page lists a desktop app, iPhone and Android apps, Search Companion and a default-search option. It also promotes access through WhatsApp, Telegram and X, where users can ask questions, generate images and use real-time search without leaving the chat (Perplexity).
For developers, the documented Search API has a narrower role: it returns real-time, ranked web results as structured data. Its controls cover domains, languages, regions and extracted content (Search API quickstart).
Suggestion: Decide first whether you want an interactive Perplexity access route or structured web results for software. That distinction is more useful than assuming every product labeled “search” produces the same type of response.

Search results and cited answers are different outputs
The Search API returns ranked entries containing titles, URLs and extracted text; it does not produce a generated answer. Perplexity directs developers who want an LLM-generated answer with citations to its Agent API instead (Search API quickstart).
This documented distinction matters when evaluating citations. A URL in raw search results identifies a result, while the cited-answer workflow belongs to the Agent API. The supplied official documentation supports citations for Agent API answers, not a broad promise that every Perplexity interface or response will include them.
Suggestion: If citations are essential to a developer workflow, evaluate the Agent API rather than expecting the Search API to compose a cited narrative. Even then, treat each citation as a route to evidence: open the linked page and check whether it supports the relevant statement. This checking method is an editorial recommendation, not a guarantee about accuracy.

Documented Search API access
Perplexity documents official SDKs for Python and TypeScript. The installation commands use the perplexityai Python package or the @perplexity-ai/perplexity_ai TypeScript package. Authentication uses a Perplexity API key, which the SDKs can read from the PERPLEXITY_API_KEY environment variable (Search API quickstart).
The same quickstart documents one Search API endpoint: POST https://api.perplexity.ai/search. A basic request supplies a query and may set max_results. The response exposes ranked results, including a title and URL, while the documentation also describes extracted text in the returned data.
The official page additionally offers an interactive Playground that can be tried without an API key. It also describes a Perplexity command-line interface that returns ranked results as JSON (Search API quickstart).
Suggestion: Use the Playground for an initial evaluation, an official SDK for an application integration, or the CLI when JSON output fits a terminal-based workflow.
Limits and implementation boundaries
For a Search API request, max_results accepts values from 1 through 20 and defaults to 10. The Search API does not accept Agent API fields such as model, effort, instructions, tools or max_output_tokens (Search API quickstart).
Because the response contains results rather than a generated answer, an application must either process those results itself, pass them to its own model, or use the Agent API with its web-search tool. These are the alternatives described by Perplexity’s documentation.
The quickstart also warns against printing any portion of an API key because terminal history, logs and coding-agent transcripts can retain it. It recommends reading the key from the environment and avoiding keys in fixtures or commit messages (Search API quickstart).
Suggestion: Validate result-count inputs, keep Search API and Agent API request fields in separate schemas, and treat key handling as part of the integration design.
Conclusion
For interactive access, consider one of the Perplexity routes listed on its official site, including its desktop and mobile options or supported messaging services (Perplexity). For software that needs ranked, structured web results, the Search API is the documented match. For an LLM-generated answer with citations, Perplexity instead points developers to the Agent API (Search API quickstart).
Practical recommendation: Start by writing down the required output—interactive response, raw ranked results or generated cited answer. Choose the corresponding access route, test it with a representative query, and inspect linked evidence before relying on important claims. For a Search API integration, begin in the Playground and then adopt an official SDK if the returned structure meets the application’s needs.
Sources
Disclosure
This article was prepared with AI assistance.
Improve this guide
Report a problem or suggest the next question
Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.
