
TL;DR
- Valyu’s Search API lets you retrieve full-text patents from USPTO, EPO, and WIPO in 3 lines of code
- Search by natural language and get structured JSON that includes title, abstract, inventors, CPC codes, citations, and full claim text
- Fully compatible with LangChain, Vercel AI SDK, or LlamaIndex
Why Patent Data Matters for AI Builders
Patents are the most concentrated, high-signal source of technical and legal innovation but they’re often underused because they’re difficult to work with.
Valyu makes it simple to build:
- Prior art search agents for research or litigation
- Technical scouts that track innovation across domains or inventors
- Citation-aware assistants that map invention timelines
- Pharma/biotech R&D copilots that combine patents with trials + papers
- Enterprise legal tools for IP monitoring, competitive strategy, or FTO analysis
The Problem With Traditional Patent Access
- Raw XML or PDF: no clean structure for LLMs
- Only keyword search: can’t ask questions semantically
- Different metadata schemas (USPTO vs WIPO vs EPO)
- No unified API for global filings
- No support for long-form answers or RAG use
Even when APIs exist, they return filing numbers, not usable context. And most systems choke on the legal formatting of claim text.
The Fast Way: Use Valyu’s Patent Search API
Valyu indexes and structures global patent data, making it searchable and RAG-ready for LLM-native applications.
- Natural language search
- Unified coverage (USPTO, EPO, WIPO)
- Full-text retrieval (title, abstract, description, claims)
- Structured metadata: inventor, assignee, CPC, jurisdiction, citations
3-Line Setup
1import { Valyu } from 'valyu-js';23const valyu = new Valyu({ apiKey: 'your-valyu-api-key' });45const response = await valyu.search(6 "battery cooling systems for electric vehicles filed after 2020"7);89console.log(response);
Get your API key
Explore the patent search docs
Example Use Cases
Prior Art Finder
“Show patents describing antibody sequence optimization in oncology.”
Patent Landscape Tool
“List patents on LLM alignment techniques filed since 2022.”
Enterprise Competitive Intel Agent
“Find recent filings by Samsung in wearable biosensors.”
Drug Discovery Assistant
“Pull patent claims related to CAR-T therapy and CD19 antigens.”
Advanced Usage Example
Query patents by jurisdiction, date range, and CPC code:
1const response = await valyu.search(2 "gene editing delivery methods with lipid nanoparticles",3 {4 included_sources: ["valyu/valyu-patents"],5 start_date: "2025-01-01",6 relevance_threshold: 0.6,7 max_num_results: 58 }9);
💡 You can also filter by assignee, inventor, or combine with PubMed/clinical trials for richer biomedical context.
Live Demo
Try the Patent Search Demo
Search international patents using plain language. Retrieve structured results with full claims, inventor names, and CPC classifications — ready for tool calls or long-context reasoning.
Best Practices for Patent Retrieval with LLMs
- Use relevance_threshold to control for overly broad matches
- Pair with academic sources for evidence synthesis (e.g., “patent + paper” view)
- Use CPC filters to target domains precisely (e.g., A61K = pharma, G06F = computing)
- Summarize or chunk claims before injecting into LLM context
- Add assignee/company names to track IP portfolios
FAQ (Schema-Enabled)
Q: What jurisdictions are covered?
A: USPTO, EPO, and WIPO filings are included with full metadata.
Q: Do you return full text of patents?
A: Yes, including title, abstract, description, and claims.
Q: Can I search by CPC, assignee, or inventor?
A: Yes, all metadata fields are filterable via structured config or in the natural language query.
Q: Are citations and prior art included?
A: Yes, backward citations (to prior filings) are included in structured fields.
Build Patent-Aware AI Apps in 3 Lines of Code
Whether you’re building an agent for pharma, hardware, or legal workflows, Valyu gives you semantic access to the global patent corpus clean, structured, and production-ready.
🔑 Get your API key
📚 Read the patent docs
🧠 Use with LangChain