What is Valyu?
AI-native search and retrieval API for LLMs and agents, querying the open web plus proprietary sources like arXiv, PubMed, SEC filings, and financial data.
Our verdict
Valyu is an AI-native search and retrieval API built for LLMs and agents rather than human readers. It markets itself around "DeepSearch" and exposes four documented endpoints: Search (web plus proprietary sources), Contents (URL extraction), Answer (grounded question answering), and DeepResearch (an async agent that returns a cited report). Results come back as structured, markdown-formatted content shaped for model context windows instead of ranked links. The differentiator versus peers like Exa and Tavily is breadth of specialized sources: alongside web search it queries arXiv, PubMed, full-text SEC filings, clinical trials, patents, and financial and market data. The Contents endpoint can do full Chrome rendering for JavaScript-heavy pages, though that is one extraction option rather than the core product.
The honest caveat is maturity and scale. Valyu is early-stage: a small team with UK roots (the legal entity is VALYU.NETWORK LTD, founded out of UCL) and a San Francisco footprint, having pivoted from an earlier data-licensing and web3 identity to the current search API. The GitHub org (github.com/valyuAI) is genuinely active with frequent pushes, and a16z named Valyu alongside Tavily in its "Search Wars Episode 2" piece, but the customer case studies it publishes (RevisionDojo, o11, Storke, Holistic AI) are vendor-written rather than independent, and there is no reliably confirmed funding figure (aggregator numbers conflict and could not be verified against a primary source, so we list none). The benchmark wins Valyu publishes on FreshQA, SimpleQA, and finance and medical retrieval are its own first-party claims and have not been independently reproduced.
Worth evaluating if you are building agents that need depth in finance, academic, or regulatory sources and want LLM-ready output from a single API. For broad, proven general-web recall at production scale, the more established peers in this category still have the longer track record. Note it is hosted only with API-key auth (no self-hosting), and only the client SDKs are open source.
Categories:
AI search APIs are the infrastructure layer that gives large language models access to current web information. Unlike traditional search engines, these APIs return semantically relevant, structured results optimized for retrieval-augmented generation (RAG) and AI agent workflows. They are used by AI products that need to answer questions about the real world beyond their training data.
