← Back to directory

Valyu

Search, built for AIs doing knowledge work.

Visit Valyu

Valyu is a search API built for AI agents and knowledge work: one API surface (Search, Contents, Answer, DeepResearch) that spans the open web plus specialist and proprietary sources - SEC filings, academic papers, clinical trials, patents, financial markets, and more - aimed at finance and science/R&D use cases with cited, structured outputs.

Vendor's own claim
other/unknown

Core capabilities

Vendor's own claim
  • Search API: multi-source search across web, academic, financial, and proprietary data sources with source filtering and relevance control
  • Contents API: extract clean, LLM-ready content from up to 10 URLs, with optional AI summaries and structured extraction
  • Answer API: grounded, cited answers via server-sent-event streaming
  • DeepResearch API: asynchronous multi-step research agent producing cited reports and deliverables (xlsx/docx/pptx/pdf/csv)
  • Workflows (Beta): templated, versioned DeepResearch for repeatable tasks like earnings previews or diligence reads
  • Datasources API: programmatic discovery of available sources and categories
  • Access to 55+ sources including SEC filings, arXiv, PubMed, clinical trials, ChEMBL, PubChem, DrugBank, FRED, World Bank, IMF, and patents

API & agent surface

From official docs
Authentication
API key passed via the X-API-Key header (lowercase x-api-key also accepted)
Base URL
https://api.valyu.ai
Notable endpoints
  • POST /v1/search - multi-source search across web and proprietary datasets
  • POST /v1/contents - extract clean, structured content from URLs
  • POST /v1/answer - AI-powered answers with real-time search
  • POST /v1/deepresearch/tasks - start an async DeepResearch task (also runs Workflows via workflow_id)
  • GET /v1/deepresearch/tasks/{id} - poll task status and retrieve the report
  • POST /v1/deepresearch/batches - run many research tasks in parallel
  • GET /v1/workflows - list Workflow templates (Beta)
  • GET /v1/datasources - discover available sources and their schemas
Machine-readable spec
OpenAPI spec retrieved

Pricing

From official docs
PlanPriceIncludedNotable
Pay As You Go (Free)$0$10 free credits on signupWeb, Arxiv, Pubmed
Ridiculously Cheap$29/month$50/month in credits (~$0.80 per credit)Web, Arxiv, Pubmed, Stock prices, Crypto, Forex, Polymarket, Kalshi, MedRxiv, BioRxiv
Comfortably Pro$89/month$130/month in credits (~$0.70 per credit)Everything in Ridiculously Cheap, plus:, SEC Filings, Insider transactions
Serious Business$449/month$750/month in credits (~$0.60 per credit)Everything in Comfortably Pro, plus:, ChEMBL, PubChem, DrugBank, OpenTargets, NPI Registry, WHO ICD, NIH Research Grants
EnterpriseContact for pricingCustom rate limits, Access to all data sources, Higher results per query

Checked 2026-08-06 against their pricing page. Vendors change pricing without notice.

What is Valyu

Vendor's own claim

Valyu positions itself as search infrastructure purpose-built for AI agents rather than humans: a single API and credit system spans four products (Search, Contents, Answer, DeepResearch) and unifies open-web search with paywalled/specialist datasets across finance and science, each retrievable with per-source pricing and citations traced to source.

Ideal customer profile

Vendor's own claim

Developers and teams building AI agents or products that need grounded, cited research across finance or scientific/R&D domains - RAG pipelines, agent tool calls, due-diligence or research-report generation - where general web search doesn't reach the specialist sources the answer depends on.

Who should look elsewhere

Our read

Pricing is entirely usage-based (CPM for Search, per-URL for Contents, fixed per-task for DeepResearch), with the free tier limited to web, arXiv, and PubMed only - every other specialist source (SEC filings, clinical trials, patents, drug-discovery data, economic indicators) requires a paid subscription starting at $29/month. Teams needing predictable flat-rate costs, or wanting proprietary data sources without committing to a metered credit model, would find the CPM-based Search pricing ($0.50-$30 by source) and DeepResearch task pricing ($0.10-$15, with the top "max" mode requiring at least $15 in available credits) harder to budget against than a flat-fee competitor.

Valyu FAQ

Does Valyu have an MCP server?

Yes.

From official docs
Is there a free tier?

Yes, $10 in free credits on signup with no card required, covering Web, arXiv, and PubMed access.

From official docs
How is Valyu priced?

Usage-based, shared credit pool ($1 = 1 credit baseline). Search is priced per-mille-tokens ($0.50-$30 by source), Contents at $0.001 per URL plus AI costs, Answer at $0.10/request plus variable costs, and DeepResearch at a fixed per-task rate ($0.10-$15 depending on mode). Monthly subscriptions ($29-$449) lower the per-credit rate and unlock specialised sources.

From official docs
What is DeepResearch and how is it different from Search?

DeepResearch is an asynchronous autonomous research agent that plans, searches, verifies facts, and writes a cited report; the docs recommend it over hand-rolling a loop of raw Search calls for research-style tasks, reserving raw Search for tight, low-latency control inside a custom agent loop.

From official docs
What SDKs does Valyu offer?

Python (pip install valyu), TypeScript (npm install valyu-js), and an Alpha-stage Rust SDK, all covering Search, Contents, Answer, DeepResearch, and Datasources.

From official docs
What frameworks does Valyu integrate with?

The docs list integrations for the Vercel AI SDK, LangChain, and LlamaIndex as tool definitions and retrievers, plus a Claude Code plugin and agent skills for Claude Code, Cursor, and Windsurf.

From official docs

Verify before you buy

  • Exact list of named MCP tool functions exposed by the hosted MCP server
  • Social media links (LinkedIn, X, YouTube, GitHub)

Other products that solve a similar job or fit the same part of the GTM stack.

Need help setting it up?

This page tells you whether Valyu fits. Cheetah builds and runs the system it goes into — enrichment, sequencing, CRM and the joins between them.

Book a call →

The directory stays free. There is nothing gated behind this.