Features
Agentic search with reasoning
1.9x better recall than embedding-only methods
24x faster than traditional embedding approaches
Test-time compute for dynamic query adaptation
RL-based search optimization with 1k+ QPS rollouts
Outperforms frontier models on complex search tasks
Training to beat GPT-5 at search (ongoing)
Designed for context-aware AI system retrieval
Waitlist access for research phase
Fetch endpoint: retrieve web page content in <2 seconds
Search endpoint: sub-second web search (standard and deep modes, 1-3s)
Research endpoint: asynchronous deep research (1-10 minutes)
Full-text snippets from trusted sources
Configurable source selection and freshness controls
Private index deployment for proprietary data
Bring your own cloud (BYOC) deployment
99.9% uptime SLA backed by enterprise agreements
92% F-score on SimpleQA factual accuracy
61% SealQA-0 score on deep research
Integration with OpenAI SDK wrapper, LangChain, CrewAI
GDPR compliant architecture
SOC 2 Type II and ZDR on every plan
Predictable pay-per-request pricing
Replace SerpAPI with single authenticated call
