Perplexity

Perplexity exposes a Search API ($5.00 per 1,000 requests, $1.00 for fast search) and an Agent API in which the model decides when to call tools. The sourcing-relevant tool is people_search ($5 per 1,000 invocations), which finds people by name, role, company, education, skill or location and returns publicly available professional details, with documented token-budget parameters and a usage policy barring doxxing and unauthorized background screening. It sits inside Agent API requests, so cost also includes model tokens, and results are model-mediated rather than a filterable candidate index. The older Sonar chat-completions endpoint carries a deprecation notice (support stated until September 27, 2026) while the migration overview still describes it as supported, which is a documentation inconsistency we could not resolve.

Documentation is organised, with an SDK overview, a Sonar-to-Agent migration guide, tool references and an error-handling guide. Official SDKs exist for Python (perplexityai) and Node.js/TypeScript. A public status page (status.perplexity.com) shows 100% API uptime for July-October 2026, though no SLA was found on the pages reviewed. Self-serve key creation and published per-request prices make onboarding straightforward, but we found no free credits and no numeric rate limits on the pages read. We did not verify an official MCP server, so none is counted. No G2/Trustpilot/Reddit developer sentiment was retrieved.