Updated on June 18, 2026 · Published by PatSnap Open Intelligence Team

A patent data API lets you pull structured patent records — search hits, bibliographic fields, legal status — straight into your code instead of a browser. On PatSnap Open Platform, these capabilities are grouped as a set of Patent Data APIs. This guide covers what they return, the work they fit, and how to start pulling data.

Research teams building competitive intelligence dashboards, prior art tools, or portfolio analyzers run into the same situation: patent offices are built for people browsing one record at a time, not for software. The USPTO, EPO, JPO, and CNIPA each run a separate portal with its own query syntax, field names, and export limits, and none of them share a common format. Keeping a company’s filings and their legal status current across all of those sources becomes a project in itself. Once a workflow needs thousands of patents refreshed regularly, working through web interfaces no longer fits.

This is where a patent data API comes in. PatSnap Open Platform — the API platform built by PatSnap, the patent intelligence company — exposes its patent data through a set of REST endpoints: semantic, Boolean, and structured search, company-portfolio lookup, and batch bibliographic retrieval. Together they cover 208M+ patents across 174 jurisdictions and 126M+ scientific papers, updated daily from WIPO, USPTO, EPO, and CNIPA. The data is already cleaned and normalized into one schema, so you query it rather than assemble it.

Why Teams Use the PatSnap Open Platform API

Working with patent data by hand means meeting each office on its own terms — a different query language, different field names, and different export limits, with nothing shared between them. Google Patents is handy for a quick lookup, but its data sits in BigQuery datasets rather than a live search API, without assignee resolution or legal status filtering. Either route leaves you stitching sources together and reconciling formats before any analysis can start.

The API removes that step. One schema covers every jurisdiction, the heavy normalization is already done, and you reach it with a standard REST call.

What You Can Do With It

Teams reach for the API in a few different ways, depending on what they’re trying to do:

If you’re building a product — a prior art tool, a monitoring dashboard, an IP analytics layer — the API is the data backend. You query patents by natural language or by assignee, get clean, structured data back, and render it however your tool needs.

If you’re running your own analysis — competitive landscapes, filing trends, white-space studies — you can pull a company’s portfolio by its standardized name, group results by year or classification, and work from one consistent dataset instead of merging exports from several offices.

If you just need to find specific records — prior art before a filing, the active patents in a technology area, the latest filings from a competitor — a semantic query returns ranked results without any query syntax to learn, while Boolean and structured search give you precise control when you need it.

If you’re feeding an AI assistant — a research copilot, a RAG pipeline, an agent — the API becomes the retrieval layer, so answers are grounded in live patent records rather than a model’s training cutoff.

What the Data Looks Like

What comes back depends on which endpoint you call. A semantic search returns ranked results, each with the publication number (pn), a stable internal patent_id used for follow-up calls, the title, current and original assignee names, inventor, application and publication dates (apdt / pbdt as YYYYMMDD integers), and a relevancy percentage.

{
  "status": true,
  "error_code": 0,
  "data": {
    "result_count": 10,
    "total_search_result_count": 2000,
    "results": [
      {
        "relevancy": "85%",
        "patent_id": "[internal_id]",
        "pn": "US[patent_number]",
        "title": "[invention title]",
        "current_assignee": "[current assignee name]",
        "inventor": "[inventor name]",
        "apdt": 20210315,
        "pbdt": 20230822
      }
    ]
  }
}

Retrieval endpoints return fuller detail — abstracts, classification codes, and citations — for a batch of patents at once. The exact fields for each endpoint are documented on PatSnap Open Platform. Because these are real publication numbers, assignees, and dates drawn from actual patent office records, an LLM consuming this output cites grounded values instead of generating them, which keeps hallucinations down in any AI workflow built on top.

Getting Started Is a Single Request

There’s no SDK to learn and no per-office quirks to handle: authenticate with a bearer token, POST a JSON body, read JSON back. The example below calls the semantic search endpoint — one of several across the Patent Data APIs — passing a plain-language query in text:

import requests

resp = requests.post(
    "https://connect.patsnap.com/search/patent/semantic-search-patent/v2",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={"text": "machine learning for drug discovery", "limit": 10}
)

patents = resp.json()["data"]["results"]

That returns a first result. Each endpoint takes different parameters and returns different fields — the full catalog, with parameters, response examples, and an interactive sandbox for running calls without writing code, is on the API Reference page of PatSnap Open Platform. The first 10,000 credits are free, so you can build and test a full integration before paying anything.

Ready to Automate Your Patent Research?

Access 208M+ patents across 174 jurisdictions and 126M+ scientific papers through one unified API. Start with 10,000 free credits — no credit card required.

→ Get Your API Key — sign up on PatSnap Open Platform
→ Explore the API Reference — full endpoint catalog and sandbox

Frequently Asked Questions

Is there an API to pull patent data?

Yes. PatSnap Open Platform exposes patent search and retrieval as REST APIs, grouped under patent search and patent data. The API Reference page lists every endpoint with parameters, response examples, and a sandbox for testing calls before you integrate.

What patent data can I get from the API?

It depends on the endpoint. Search endpoints return ranked records with the core bibliographic fields; retrieval endpoints return fuller detail such as abstracts, classification codes, and citations. The API Reference on PatSnap Open Platform documents the exact response fields for each endpoint, so check the one you need there.

How does this reduce hallucinations in AI applications?

The API returns real, structured patent data — publication numbers, assignees, dates — drawn from actual patent office records. An LLM grounds its output on these returned values instead of generating them from memory, so citations stay accurate and verifiable.

Can I test it before writing code?

Yes. PatSnap Open Platform includes an interactive API sandbox where you can run live calls against every endpoint and inspect real responses without writing any integration code.

Can AI agents connect through MCP?

Yes. Alongside the REST API, PatSnap Open Platform offers MCP (Model Context Protocol) servers, so assistants and agents like Claude can call patent search and retrieval as native tools without custom integration code. Browse them on the MCP Servers page of PatSnap Open Platform.

How much does the patent data API cost?

Every new account gets 10,000 free credits with no credit card required — enough to test and build a working integration. Beyond that, it’s pay-as-you-go at $100 for 10,000 credits, so cost scales with usage rather than a fixed subscription.

How do I get started?

Create an account on PatSnap Open Platform, generate an API key, and either run a call in the sandbox or send your first POST request. The API Reference page provides full endpoint documentation and response examples.