🔬 Patent Lookup: a company's patents, an inventor's patents, prior art, one document in full

Find out what a company has patented, what a person has invented, or what has already been filed on an idea, from the European Patent Office's official worldwide patent data. Five modes, one Actor: a company portfolio with every row's applicant checked against the company name, an inventor's patents, a keyword search with CPC and office filters, a single publication number opened with its family, legal events, citations and claims, or every member of a patent family. Try it now with the prefilled input: the portfolios of Mobileye Vision Technologies Ltd and Iscar Ltd.

$0.002 a search and $0.0015 a patent row, for investors and analysts sizing a patent moat, competitive intelligence teams, R&D and founders running a prior-art check, recruiters and IP operations. This patent search API does in one call what you would otherwise do by hand in Google Patents or Espacenet, one query and one company at a time, and it returns the rows as JSON with the family, the legal status and the citations attached.

The data is the EPO's own worldwide bibliographic database (DOCDB, 100+ patent offices), its legal-events database (INPADOC) and its full-text collections, read through the EPO's Open Patent Services under the EPO's terms of use. No scraping of Google Patents, no data broker, no inventor profiles built from anywhere else.

What you get

• ✅ Company portfolios that keep namesakes out. A company query is not a text search for the name. Every row that comes back has its applicant compared with the company name you gave, so "Medical Laser Systems" does not fill up with "Topcon Medical Laser Systems". Each row says whether the applicant is the company itself, another party, or unknown. Frequently requested companies answer from a pre-built portfolio that is refreshed every 90 days and stamped with its as_of date.
• ✅ Every office, one schema. US, EP, WO, CN, JP, KR, IL, DE, GB and the rest of the EPO's coverage return the same fields: publication number, title, abstract, inventors, applicant, priority, filing and publication dates, estimated expiry, granted flag, family id, IPC, CPC, Google Patents and Espacenet links.
• ✅ One document in full. Family members across offices, INPADOC legal events (examination, grant, fees, lapse), forward citations, and the claims and description text where the EPO carries it (EP and WO publications and several national collections).
• ✅ Priced per row. $0.002 per search, $0.0015 per patent row, $0.0075 per document, $0.003 per family, plus $0.00005 to start the run. A 50-patent portfolio costs about 8 cents. No match and invalid input cost nothing.

This is not a legal opinion and not a certified register extract. For legal proof order the document from the patent office.

👥 Who uses a patent lookup?

• 📈 Investors and analysts. Read a startup's patent moat before a round: how many families, which offices, granted or pending, who the inventors are, and whether the filings are recent, without a namesake's patents landing in the count.
• 🏢 Competitive intelligence. Pull a competitor's filings by office and year, collapse them to one row per family, and schedule the same run monthly to see what is new.
• 🧑‍🔬 R&D and founders. Prior-art check before you file: keywords in title and abstract, narrowed by CPC class, office and publication date, with claims text on the EP and WO documents you want to read closely.
• 🎓 Recruiters and researchers. An inventor's patents across employers, plus the companies whose portfolios name that inventor, so a career reads from the patent record.
• ⚖️ IP and legal operations. A publication number opened with its family, its legal events and its forward citations, in structured JSON, for docketing checks and freedom-to-operate triage.
• 🤖 AI agents and pipelines. Structured JSON with the source and the as-of date on every row, callable from the Apify API, the MCP server and the Python and JavaScript clients. Priced per event, the platform's $0.00005 Actor start included, so it is eligible for agentic payment.

⬇️ Input: a mode and a list of queries

One run, one mode, any number of queries read in that mode. The prefilled input runs two company portfolios.

{
  "mode": "company",
  "queries": ["Mobileye Vision Technologies Ltd", "Iscar Ltd"],
  "max_results": 100
}

Field	What it does
mode	company (patents the company owns), inventor (patents naming a person), search (keywords with filters), document (one publication number in full), family (every member of a publication's family).
queries	One query per line: company names, inventor names, search phrases or publication numbers. Each entry may also be an object with query, mode and any filter below, so one run can mix modes.
max_results	Rows per company, inventor or search query, up to 500. Each row is one patent_record event.
countries	Offices to keep (US, EP, WO, CN, JP, KR, IL and the other two-letter codes).
granted_only, one_per_family	Keep granted patents only. Collapse each family to one row.
published_from, published_to	Inventor and search modes, publication date as 2020, 2020-06 or 2020-06-15.
filed_from, filed_to	Company mode, filing date as YYYY-MM-DD.
cpc, ipc, applicant, inventor	Search mode: classification symbols (comma separated, any matches) and applicant or inventor names to narrow the keyword search.
company	Inventor mode: keep patents where this company is the applicant.
include	Document mode: family, legal, citations, claims, description. Default family, legal and citations.
offset	Company, inventor and search modes: skip this many results before the first row, so a second run pages past the first max_results instead of returning them again.
refresh	Company mode: query the EPO now even when a pre-built portfolio exists.

An empty query, a query over 200 characters (300 for a search), an unknown mode, or a document or family query that is not shaped like a publication number is rejected before any paid step, as an uncharged error item, and the run still ends normally.

Publication numbers are accepted in every common form: EP3000000A1, US12724156B1, WO2023187781A1, US 2024/0394598 A1, ep3000000.

⬆️ Output: one dataset item per patent row

Every item carries type, the query and mode it came from, and charged, so the dataset is the complete record of what you paid for.

type	Charged	Contents
patent	once per row (patent_record)	one patent row from a company, inventor or search query
search_summary	once per company, inventor or search query (patent_search), once per family query (patent_family)	match status, source and as-of date, the company resolved, candidates, related companies, totals, has_more, truncated
document	once (patent_document)	the publication with the sections you asked for
family_member	never (the family's summary item is the charged one)	one member of the family, full patent row
error	never	an error code and a message

A patent row from the first prefilled query:

{
  "type": "patent", "query": "Mobileye Vision Technologies Ltd", "mode": "company",
  "charged": true, "charged_event": "patent_record",
  "publication_number": "US20260099146A1", "country_code": "US", "patent_number": "20260099146", "kind_code": "A1",
  "family_id": "93936260", "document_type": "publication", "application_id": "US202519416498",
  "priority_date": "2023-06-23", "filing_date": "2025-12-11", "publication_date": "2026-04-09",
  "estimated_expiration": "2045-12-11", "is_granted": false,
  "title": "Systems and methods for vehicle navigation",
  "abstract": "A system for navigating a host vehicle relative to a road segment is disclosed. The system may comprise at least one pro ...",
  "inventors": ["Lotan Roy Maor", "Harouche Shahar", "Shaag Nadav"],
  "applicant": "Mobileye Vision Technologies Ltd", "applicant_country": null, "applicant_match": "company",
  "priority_country": null, "ipc": "G05D1/243",
  "cpc": ["B60W10/20", "G05D1/2435", "G06N20/00", "G06N3/045", "G06V10/764", "G06V10/82", "G06V20/58", "G06V20/582", "G06V20/588"],
  "family_size": 100,
  "google_patents_url": "https://patents.google.com/patent/US20260099146A1/en",
  "espacenet_url": "https://worldwide.espacenet.com/patent/search?q=pn%3DUS20260099146A1",
  "source": "stored", "as_of": "2026-08-14", "search_term_used": "MOBILEYE VISION TECHNOLOGIES LTD", "confidence": 1.0
}

Its summary item, showing the company the query resolved to and how much more there is:

{
  "type": "search_summary", "query": "Mobileye Vision Technologies Ltd", "mode": "company",
  "charged": true, "charged_event": "patent_search", "status": "matched", "source": "stored",
  "organization_name": "Mobileye Vision Technologies Ltd.", "organization_as_of": "2026-08-14",
  "organization_patents_total": 290, "candidates": [], "related_companies": [],
  "n_records": 100, "total_available": 290, "has_more": true, "truncated": false,
  "search_term_used": "MOBILEYE VISION TECHNOLOGIES LTD", "cql": null, "elapsed_s": 3.1
}

source is stored when the answer comes from a pre-built portfolio (then as_of is the date of its last refresh, at most 90 days ago) and live when the EPO was queried at the moment of your run (then as_of is today). applicant_match says whether the row's applicant is the company itself, another party (a co-applicant, a licensee, an inventor filing in their own name, or a transliterated name) or unknown. related_companies on an inventor query lists companies whose patent portfolios name that inventor, with their filing years.

A document item carries, on top of the patent row, family (members with office, kind and date), legal_events (INPADOC codes with dates and descriptions, for the whole family), citations (forward_count and the citing publications, backward citations where the EPO lists them), and claims and description as {lang, text} where the EPO carries full text.

Export the dataset as JSON, CSV, Excel or XML, or read it through the Apify API. Lists (inventors, cpc, legal_events) are kept as JSON in CSV exports.

💰 How much does a patent lookup cost?

Event	Price	When it fires
patent_search	$0.002	once per company, inventor or search query, whatever it finds
patent_record	$0.0015	once per patent row returned by such a query
patent_document	$0.0075	once per publication opened in document mode, when it exists
patent_family	$0.003	once per family query that found the family
Actor start	$0.00005	once per run
No match, invalid input, rate limit, upstream failure	free

The worked examples below are the event charges only. The Actor start adds $0.00005 a run, so the prefilled input is $0.304 of events and $0.30405 in total:

• The prefilled input, two portfolios at 100 rows each: 2 x $0.002 + 200 x $0.0015 = $0.304.
• A 50-patent startup portfolio: $0.002 + 50 x $0.0015 = $0.077.
• One inventor with 12 patents: $0.020.
• A prior-art search returning 500 rows: $0.752.
• One publication opened with family, legal events and claims: $0.0075.
• A family of 8 members: $0.003.
• 1,000 company names checked, 600 with patents averaging 20 rows: 1,000 x $0.002 + 12,000 x $0.0015 = $20.00.

Set a spending limit on the run for a hard cap: the Actor stops before the next query once it is reached and marks any row it could not charge.

🗺️ What is covered

• Offices. The EPO's worldwide bibliographic database covers publications from more than 100 patent offices, with English titles and abstracts where the office supplies them. Claims and description text are available for EP and WO publications and several national collections. Legal events come from INPADOC and cover the offices that report to it.
• Companies. Any company that has filed a patent anywhere in the EPO's coverage. Frequently requested companies answer from a pre-built portfolio, built from the company's registered legal names (current, alternate and past) with only the families whose applicant matches one of them kept, refreshed every 90 days and dated. Every other company is searched live at the EPO with the same applicant check.
• Inventors, keywords, documents and families are always answered live from the EPO at the moment of the run.

🔎 How the lookup works

1. A company name is matched against the pre-built portfolios first: an exact match on the registered names, then a close match on the legal name. One clear winner is served from its portfolio with the as_of date. Two strong candidates come back as ambiguous with the shortlist, uncharged rows. Otherwise the EPO is searched live for the name as applicant, and every row's applicant is checked against the name before it is returned.
2. An inventor name, a keyword search and a document number are sent to the EPO as one query in its own query language (the cql field in the summary shows it). Filters for office and publication date go into the query, the granted and one-per-family filters are applied to what comes back.
3. Rows are resolved from the EPO's bibliographic service in batches, under a wall-clock budget of 150 seconds per query. A query cut short by the budget says so (truncated: true), and has_more on the summary says there are more rows: run it again with offset set to what you already have.
4. A document is opened by its exact publication number (the kind code decides the document, without it the granted publication wins, else the latest), then the sections you asked for are fetched one by one.
5. Nothing is estimated or inferred and no language model is involved: every field is the EPO's own value, and confidence scores the company-name match only.

⏱️ How long does a query take?

A pre-built portfolio answers in 2 to 4 seconds. A live company, inventor or search query takes 3 to 10 seconds for 100 rows and up to two minutes for 500, because the EPO paces search and retrieval per account. A document with family, legal events and claims takes 5 to 10 seconds. Queries run one after another, and two live queries may be in flight per API key, so a run of 100 company names finishes in a few minutes to a quarter of an hour.

🔌 Patent API, clients and integrations

Run the Actor from the Apify Console, from the Apify API, from the Python or JavaScript client, from the Apify MCP server, or from Make, Zapier and n8n through the Apify integrations. Schedule it to re-run a competitor list monthly, and set a webhook on run completion to pull the dataset into your own system.

from apify_client import ApifyClient

client = ApifyClient("<YOUR_API_TOKEN>")
run = client.actor("pvalyou/patent-lookup").call(run_input={
    "mode": "company",
    "queries": ["Mobileye Vision Technologies Ltd", {"query": "EP3000000A1", "mode": "document", "include": ["family", "legal", "claims"]}],
    "max_results": 50,
})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item["type"], item.get("publication_number"), item.get("title"), item.get("applicant"), item.get("charged"))

🔐 Data, licences and privacy

• Official source. Every answer is read from the European Patent Office's Open Patent Services, under the EPO's terms and conditions for OPS, which allow OPS data to be included in our own products and services and distributed as part of them. We do not scrape Google Patents or any other patent website, and we do not redistribute the EPO's data set as such: there is no bulk export mode.
• Not a certified extract. Bibliographic data, legal events and full text are the EPO's records as published. For legal proof order the document from the patent office. Legal events are informational and may lag the national office.
• What we keep. Pre-built company portfolios are retained, refreshed every 90 days and served with their as_of date, so you know how old such an answer is (refresh forces a live pass). Every other answer is passed through from the EPO to your run and not retained. Operational logs keep the query text, the mode, timing and the billing outcome, which is what running and billing the service needs.
• Personal data. Inventor names are printed in every patent publication by law, and we return them as the EPO publishes them. We build no inventor profiles from other sources. related_companies on an inventor query is computed from the patent record alone. Use inventor data in line with the laws of your jurisdiction. A person who wants a patent record corrected should ask the office that published it. Objections about this Actor's processing can be raised through the Issues tab on this page.
• No access controls bypassed. No CAPTCHA solving, no logins, no credentials of yours. The EPO's fair-use rules on volume and pacing are enforced in the service itself.

❓ FAQ

Is the data live?

Pre-built company portfolios are dated (source: stored, as_of the last refresh, at most 90 days old, refresh: true forces a live pass). Everything else is live from the EPO at the moment of the run (source: live).

Why did a company come back ambiguous?

Two or more pre-built portfolios carry that name, or a name that close. The summary item lists the candidates with their number of patents. Send the exact legal name, or use search mode with applicant for a raw applicant search that is not verified.

How is the company mode different from searching the applicant field?

An applicant search on any patent site matches the tokens of the name, so "Medical Laser Systems" also returns "Topcon Medical Laser Systems". Company mode checks every row's applicant against the company's name before returning it, and pre-built portfolios were built the same way against the company's registered names. applicant_match on each row tells you what the check found.

What does applicant_match: other mean?

The row belongs to a family that was verified by its applicant, but this member of it names another party: a co-applicant, a licensee, a subsidiary under a transliterated name, or an inventor filing in their own name. The row is kept and labelled rather than hidden.

Can I get backward citations?

Forward citations (publications citing this one) are returned on every document. Backward citations (what this document cites) are returned where the EPO lists them in the bibliographic record, which is mostly EP search reports, and backward_count is 0 otherwise. This is a limit of the source, not a paid extra.

Which documents have claims text?

EP and WO publications and several national collections the EPO holds in full text. For others claims is null and claims_note says so, and the row still carries the Google Patents and Espacenet links where the full text can be read.

Is it legal to use?

Yes. The EPO's terms for Open Patent Services permit including the data in our own products and services and distributing it as part of them, with no royalties. Patent publications are public records by design. Inventor names are personal data in some jurisdictions: use them lawfully.

Is there a free tier or a trial?

Every no-match, every rejected input and every failed query is free, and Apify's free plan credit covers the first paid lookups. There is no subscription: you pay per search, per row, per document and per family, and nothing else.

What is the EPO weekly allowance?

The EPO grants a weekly data allowance per account. If paid traffic ever uses it up, live queries answer with an uncharged error until Monday 00:00 GMT while pre-built portfolios keep answering, and the allowance is raised on our side.

📏 Limits

• A query is at most 200 characters (300 for a search), and each query has a wall-clock budget of 150 seconds in the service and 180 in the Actor.
• A company, inventor or search query returns at most 500 rows per run, from the first 500 results of the EPO's ranking. The summary says has_more when there are more.
• A family answer lists at most 100 members. Legal events cover the whole family of the document opened.
• At most two live queries are in flight per API key. Queries run one after another inside a run.
• Exact duplicates (same text, same mode) within one run are collapsed. The run's spending limit is honoured.

🔗 Works well with

A patent portfolio is one answer about a company, and the other five are sold the same way: Company Enrichment API for firmographics, classification and funding, Company Registry Lookup for the registered legal names an applicant search should be run against, LinkedIn Person Enrichment for the career record of an inventor, Job Postings for what the company is hiring for now, and Website Liveness Check for whether it still has a live site.

💬 Support and feedback

Found a company that resolved wrong, a document that did not open, or a source you need? Open an issue on this Actor's page with the query, the mode and what you expected. Issues are answered by the developer.

📝 Changelog

• 0.1 (2026-09-08) First public release: company, inventor, search, document and family modes on the EPO's Open Patent Services, four priced events.