Patent data for AI, without hallucinationsWhen an AI invents plausible-sounding facts that aren't trueQuery in any language, get patents from 5 jurisdictionsThe fastest patent search in the world
findip /faɪn.dɪp/ fine-deep — Find your IP, deeply.
Problem
LLMs don't know patents
Ask GPT or Claude about a patent and they invent publication numbers that don't exist. Patent full text isn't in their training data.
Building RAG yourself means collecting tens of millions of patents, embedding them, and standing up a vector DB — months of work.
Solution
One line of findip API
We've pre-built 25 years of patents from 5 jurisdictions as vector embeddings. One API call returns semantic search results.
Connect the MCP server and Claude or GPT will answer from real patent data. No hallucinations.
Why findip
Skip building patent search infrastructure. Add patent search to your AI service with a single API.
Semantic search
Vector search by meaning, not keyword match. Searching "battery thermal management" surfaces "cooling system" and "heat dissipation" patents too.
Multilingual queries
Search in Korean, English, Japanese, or Chinese. Natural-language queries match across patents in all four languages.
MCP native
Connect Claude, ChatGPT and other AIs — they search patents directly.
Zero hallucinations
Returns only patents that actually exist, instead of LLM-invented publication numbers.
Simple API
One REST call. No vector DB to build, no embedding model to manage.
5 jurisdictions, 25 years
KR, US, JP, CN, EP — instantly search tens of millions of patents going back to 2000.
Three ways to connect
Bring findip into your AI workflow
AI integration
Add findip in Claude or ChatGPT settings and the AI searches patents directly.
MCP protocol · one-line URL
Search in the browser
Search patents in your browser with no install — sign up and start.
Playground · works right after signup
Embed in your product
Embed patent search in your own service via the REST API.
REST API · works in any language