Free trademark search API: USPTO and global registers

Every endpoint on this page was live-probed on 2026-07-29 by the Aidenix Atlas catalog.

What is a trademark search API?

A trademark search API is an endpoint that returns official mark registrations — wordmark, owner, status, classes — straight from an intellectual-property office, and 2 free routes cover most of the need. In our data — the Aidenix Atlas catalog, both endpoints live-probed on 2026-07-28 — the same test query «STRIPE» returns 537 records from the USPTO full-text search and 5,837 records from TMview, the EUIPO-run search across national registers. What the free tier honestly gives you:

• Official registration records from the offices themselves, not scraped aggregator copies.
• Search by mark text, owner name, or registration number — no API key on either route.
• Live/dead status and Nice classes, which turn a brand list into an industry profile.

Key facts at a glance

A key-facts table is the 60-second version of this page: 2 keyless routes, one US and one global — every row from our own probe run of 2026-07-28, according to the Aidenix Atlas catalog logs:

USPTO vs TMview: which route answers what?

A trademark route is a single POST endpoint with its own body dialect, and the two free ones differ in more than geography. Both answered HTTP 200 in our probe run of 2026-07-28:

Which route to use follows from the question:

• If you need US legal certainty → USPTO; it is the register of record, and ownerName carries the incorporating entity type and state.
• If you need to know whether a name is taken anywhere → TMview; the same query returned roughly 11× more records than the US register alone.
• If you are matching marks to a company → USPTO first, because its owner strings resolve to legal entities; TMview applicant names arrive transliterated across alphabets.

How do you search USPTO trademarks by owner name?

The USPTO search endpoint is a raw Elasticsearch proxy: one POST with a standard ES-DSL body, no key, no signup. According to our probe of 2026-07-28, a query_string clause searches across wordmark, ownerName, and goodsAndServices at once — which makes owner-name search a one-liner:

# note the envelope quirk: you ask for _source, the response returns it as source
curl -X POST "https://tmsearch.uspto.gov/prod-v1-0-0/tmsearch" \
  -H "Content-Type: application/json" \
  -d '{"query":{"bool":{"should":[{"term":{"WM":{"value":"stripe"}}}],"minimum_should_match":1}},"size":10,"_source":["wordmark","ownerName","alive","serialNumber","registrationNumber"]}'

The response is an Elasticsearch hits envelope: hits.totalValue for the count, then wordmark, ownerName, alive, serial and registration numbers per record. Full field list and query patterns: the USPTO access card.

We ran that query for «stripe» on 2026-07-28 — 537 US records, and here are the top 3 exactly as returned:

Read what the sample teaches, because it is the trap of this whole dataset: the payments company you were thinking of is nowhere in the top results, and all three top hits are dead marks — alive: false. Relevance ranking here scores text, not commercial reality, and expired registrations stay in the index forever. Any serious query filters on alive and reads the owner string, which conveniently carries entity type and state of incorporation — that is your join key back to a state registry.

How do you search trademarks globally?

Global search is one POST to TMview, the cross-register run by the EU Intellectual Property Office: a single basicSearch query hits USPTO, EUIPO, WIPO and dozens of national offices at once. In our probe of 2026-07-28, «STRIPE» returned 5,837 records worldwide against 537 in the US federal register alone — the global picture is roughly 11 times wider. The mechanics, according to the catalog recipe:

• Body: {"basicSearch": "<name>", "page": "1", "pageSize": "30", "criteria": "C"} — no auth.
• Response: tradeMarks array plus totalResults and totalPages for paging, 30 records per call.
• Each record names the IP office, applicant, Nice class, status, and application number — details on the TMview access card.

Our 2026-07-28 run for «stripe» returned 5,837 records worldwide, and the applicant names show the second trap of trademark data:

All three are the same US company, filed through Chinese, Arabic and Georgian offices, each register storing the applicant in its own script. Exact-string matching on applicant name fails silently here — the fix is to match on the mark plus the office plus a registry identifier, never on the name alone. This is the single most expensive lesson in cross-border IP data, and no vendor dataset removes it; they just hide it behind a resolved «company» field you cannot audit.

What do trademark records tell you about a company?

A trademark portfolio is a self-declared product roadmap: companies file marks for the names they intend to sell under, classed into the 45 Nice categories that describe goods and services. In our data, trademark records join company research from 3 angles:

• Owner-name search turns a legal entity into its brand list — including brands the website never mentions.
• Nice classes map each brand to an industry segment, filed by the company itself.
• Dead marks and fresh applications are timing signals: a rebrand, a new product line, or an abandoned market.

The join path runs both directions: resolve a company in a state registry or SEC EDGAR first, then pull its marks — or start from a mark's owner name and walk back to the registry record.

That walk-back is where the owner string earns its keep. It carries entity type and state of incorporation, so it routes you straight to the register that holds the filing: New York, Colorado and Connecticut all answer by owner name over an open API, and the company data API guide covers the routes for the states that do not. If the owner turns out to be a private issuer rather than an operating company, its financing history sits in SEC Form D filings.

Frequently asked questions

Is there a free trademark API without registration? Yes — both routes on this page are keyless. USPTO's tmsearch and EUIPO's TMview both accept anonymous POST requests; the only client requirement we found is that TMview drops connections from non-browser User-Agent strings, which reads as a dead endpoint until you set one.

Do these APIs include expired trademarks? Yes, by default, and that is a feature once you know it: dead marks reveal abandoned products and past rebrands. Our live sample above returned three expired marks in the top three hits — always filter on the status field unless history is what you want.

Can I check trademark availability with these APIs? You can check whether identical or similar marks exist, which is the factual half of the question. The legal half — whether your mark would survive examination and opposition in a given class — is an attorney's judgment, not an API response, and no free or paid endpoint answers it.

How we verify these claims

Our verification method is a live probe, not a citation chain — every count and sample on this page comes from requests we sent ourselves on 2026-07-28. What the probe run covers, in our data:

• Endpoint liveness: both routes answered HTTP 200, out of the nearly 4,000 sources the Aidenix Atlas catalog tracks.
• Live samples: the 537 US records, 5,837 worldwide records, and both result tables are unedited API output.
• Client quirks: the TMview User-Agent requirement was found by our probe failing first, then being fixed — the failure is documented above rather than hidden.
• Freshness contract: when a source drifts, the probe fleet flags it and this page gets re-verified — the «last probed» date above is that promise.

Source facts from the Aidenix Atlas catalog, probed 2026-07-28. Found an error? The catalog re-probes on every report.