NASA Patent Catalog Lookup

Purpose

Look up patents in NASA's Technology Transfer (T2) Portal — the public catalog of NASA-owned patents available for licensing. Given a free-text query, a category, a NASA center, or a specific reference ID, return one or more patent records with title, subtitle, abstract, technology description, benefits, applications, NASA case numbers, granted USPTO patent number(s), originating NASA center, technology readiness level (TRL), figure captions, and image URLs. Read-only — the skill never initiates a license application or contact submission.

When to Use

• "Find NASA patents about hall-effect thrusters / regolith 3D printing / cryogenic insulation."
• "List every JPL patent in the propulsion category."
• "Pull the full record for MSC-TOPS-89."
• Any catalog/discovery flow that needs structured NASA T2 patent data: tech-transfer scouting, prior-art surveys, commercialization research, federal-IP dashboards.
• Use the spinoff / software catalogs (/search/multi/aw/spinoff/..., /search/multi/aw/software/...) for the sibling endpoints if the user asks about NASA software releases or commercial spinoffs instead of patents.

Workflow

technology.nasa.gov exposes an undocumented JSON endpoint behind the "View More" pagination button on /patents/category/*. It returns Elasticsearch-style hits with a fully-populated _source object — every field the public detail page renders is there, plus a few that aren't (e.g. raw push_date epoch-ms, id, internal subcategory). The site is on Drupal 10 behind CloudFront with no WAF, no bot challenge, and no rate-limit-on-bare-IPs we could detect; bare Browserbase Fetch returns 200 in ~0.5s per request. A residential proxy and Verified flags are NOT required. Lead with the API; only fall back to HTML scraping if the API ever 5xx's. The site is publicly licensed federal data — there is no auth, cookie, CSRF, or session state.

1. Free-text / topical search (the primary path)

GET https://technology.nasa.gov/searchosapicat/multi/{center}/patent/{category}/{page}/{page_size}/{query}

• {center} — aw (agencywide; default), arc Ames, dfrc Armstrong, grc Glenn, gsfc Goddard, jpl Jet Propulsion Lab, jsc Johnson, ksc Kennedy, larc Langley, msfc Marshall, ssc Stennis, faa Federal Aviation Administration (yes, FAA shares the catalog).
• {category} — either all, or one of the 15 known slugs: aerospace, communications, electrical_and_electronics, environment, health_medicine_and_biotechnology, information_technology_and_software, instrumentation, manufacturing, materials_and_coatings, mechanical_and_fluid_systems, optics, power_generation_and_storage, propulsion, robotics_automation_and_control, sensors.
• {page} — 1-based page number. Empty page beyond last returns [].
• {page_size} — practical max is ~50. Above that the response can exceed 1 MB and trip Browserbase Fetch's response cap (the upstream itself is happy with larger values; only the proxy hop fails). Use 9 to match the UI default, or 50 for batched scans, and paginate.
• {query} — URL-encoded free-text. Empty (trailing slash) returns the corpus.

Response is a JSON array of {_index, _id, _score, _source} hits sorted by Elasticsearch relevance score (descending — not by date). The _source object on each hit is the full patent record:

2. Direct lookup by NASA reference ID

The JSON API does fuzzy matching on client_record_id, so passing LEW-TOPS-34 returns the 9 most-relevant LEW-TOPS-* records — not the exact one. For a known reference ID, fetch the detail HTML directly:

GET https://technology.nasa.gov/patent/{REFERENCE_ID}

• 200 → detail page rendered. Parse server-side fields directly from the HTML (selectors below).
• 302 Location: /404 → ID does not exist. Return success: false, reason: "not_found".

HTML extraction selectors (no JS execution required — Drupal renders server-side):

3. List by category (no query, no center)

GET https://technology.nasa.gov/searchosapicat/multi/aw/patent/{category}/{page}/9/

Same response shape. Use empty query (trailing slash with nothing after the page-size). Paginate until the array length is less than {page_size} — there is no totalCount field in the response, only the page itself.

4. List by NASA center (no query, all categories)

GET https://technology.nasa.gov/searchosapicat/multi/{center}/patent/all/{page}/9/

5. Lookup by USPTO patent number

Not directly supported by the API. The patent_number field stores formatted strings like "7,624,566; 10,273,944" and is not tokenized for digit-only search. Searches like /searchosapicat/.../7624566 return low-score fuzzy garbage (none of the hits actually have that USPTO number on page 1). Searches with comma-formatted values like 7%2C624%2C566 return the literal empty object {}.

Three workable patterns:

1. Scan and filter client-side. Page through /searchosapicat/multi/aw/patent/all/{page}/50/ and match _source.patent_number for the target USPTO number as a substring (after stripping commas and whitespace). The full catalog is on the order of a few thousand records — at 50/page that is ~80 round-trips.
2. Pre-build an index from the agency feed if the agent is going to do many USPTO-number lookups. Cache {usptoNumber → referenceNumber} from a one-time full crawl.
3. Use the USPTO public bulk-data API directly (ppubs.uspto.gov) — the T2 portal links out to it anyway via the <a href="https://ppubs.uspto.gov/..."> on each detail page.

Browser fallback

Use only if the JSON API ever returns 5xx (it did not during testing). All page-rendered surfaces are server-side HTML — no JS execution required:

• Search page: https://technology.nasa.gov/search/multi/{center}/patent/9/{URL-encoded query}. Response is HTML with <div class="result"> blocks. Inside each: <div class="category">...</div>, <div class="title">...</div>, <div class="description">...</div>, and <a href="/patent/{REFERENCE_ID}">Read more</a>. The 5th URL segment (9) is a slot-id placeholder — does NOT control page size or page number, and changing it returns the same 12 results. There is no HTML pagination on the search page.
• Category browse: https://technology.nasa.gov/patents/category/{category}/{center}/{page}/ — same <div class="result"> structure, with real /{page}/ pagination via the URL.
• Detail page: https://technology.nasa.gov/patent/{REFERENCE_ID} — same selectors as step 2.

The browser fallback adds no value over the JSON API for any of the discovery paths; it only exists to keep the skill resilient if NASA disables the undocumented /searchosapicat/ endpoint without notice.

Site-Specific Gotchas

• The searchosapicat endpoint is undocumented. It is invoked client-side from the "View More" button JavaScript on /patents/category/* and is not mentioned in NASA's published API docs (there are none for this portal). It could disappear without notice. If it 404s or starts returning HTML instead of JSON, fall back to scraping the server-rendered <div class="result"> blocks on /search/multi/... and /patents/category/....
• page_size is bounded by the Browserbase Fetch 1 MB cap, not by NASA. Upstream returns large pages happily, but browse cloud fetch truncates with 502 The response body exceeded the maximum allowed size of 1MB once a single page exceeds ~1 MB. Empirically: page_size=50 is safe for any single category; page_size=99 works for narrow categories (e.g. propulsion has 15 records total) but fails for all. When in doubt, use page_size=9 and paginate.
• Direct ID search via the API does fuzzy match, not exact lookup. /searchosapicat/.../9/LEW-TOPS-34 returns 9 high-scoring LEW-TOPS-* records but LEW-TOPS-34 itself is not guaranteed to be on the first page. For exact-ID lookup, always use the HTML detail page /patent/{ID} — it 200s on a hit and 302s to /404 on a miss.
• USPTO-number search is broken. Numbers like 7624566 (digits only) and 7%2C624%2C566 (URL-encoded with commas) both produce useless results. The patent_number field is stored as a single formatted string with commas and semicolons, and is not tokenized for digit lookup. Treat the API as a topic/keyword/reference-number search engine only.
• Default sort is relevance, not date. _score (Elasticsearch BM25-ish) is the descending sort key. To get newest-first, sort the page client-side by _source.push_date (which is epoch milliseconds, not seconds — divide by 1000 if comparing against Date.now() in seconds).
• reference_number vs case_number vs patent_number are three different things.
  • reference_number (a.k.a. client_record_id): the NASA T2 "TOPS" ID — e.g. LEW-TOPS-34, MSC-TOPS-89, TOP2-321. This is the URL slug on /patent/{ID}.
  • case_numbers[].case_number: internal NASA tracking IDs — e.g. LEW-19121-1, MSC-26347-1. One patent may wrap several cases. Not usable as URL slugs.
  • patent_number: the granted USPTO patent number(s) — semicolon-separated, comma-formatted, often empty for patent-pending entries. Each value links to ppubs.uspto.gov/pubwebapp/external.html?q=({num}).pn.&db=USPAT,USOCR,US-PGPUB.
• /patent/{BOGUS_ID} returns 302 Location: /404, not 404 Not Found. Configure the HTTP client to follow at most 1 redirect, then check for final.url === ".../404" or final.status === 200 — relying on the original 302 status alone is enough if you don't follow it.
• robots.txt disallows /search/ but does not mention /searchosapicat/ or /patents/category/. Respect a polite rate of ≤ 1 req/s. The site does not appear to enforce a hard rate limit (no challenges or 429s observed across ~30 fetches in the testing window), but be courteous.
• Image URLs require following a redirect. All img* / fig* URLs in the response point to https://technology.nasa.gov/t2media/tops/img/{ID}/... which 301s to https://ntts-prod.s3.amazonaws.com/t2p/prod/t2media/tops/img/{ID}/.... If you need the binary, fetch the redirect target. Most agents only need the URL, not the bytes.
• Body text may contain HTML entities and curly-quote characters. Sample observed: NASA's (apostrophe), & (ampersand), ’ (U+2019), – (en-dash). HTML-decode before display or downstream NLP.
• Center codes in _source.center are mixed-case and don't always match the URL center code. URL uses jpl, grc, larc, msfc, …; the field uses JPL, GRC, LaRC, MSFC, …. Also the TOPS-ID prefix maps differently: LEW- is GRC, MFS- is MSFC, MSC- is JSC, LAR- is LaRC, NPO- is JPL, DRC- is DFRC, KSC- is KSC, GSC- is GSFC, and TOP2- is a multi-center / agencywide marker. Don't assume the prefix in the reference number tells you the URL center scope to use.
• The corpus also contains non-TOPS reference IDs (e.g. US11847923B2 — appears to be a directly-imported USPTO ID, not assigned a NASA TOPS number). These follow the same record shape and the same /patent/{ID} URL pattern. Do not regex-filter to ^[A-Z]+-TOPS- — you'll silently drop ~1–2 % of the catalog.
• /patents/category/{cat}/{center}/{page}/'s server-rendered HTML only contains the first 9 results regardless of {page} value. Pagination is implemented client-side via the JS "View More" button — that JS is what actually hits the searchosapicat endpoint. Don't bother changing {page} in the URL; either scrape the first 9 from the HTML or (better) call the API directly.

Expected Output

Per-patent record (canonical shape, returned by both API and HTML paths)

{
  "reference_number": "MSC-TOPS-89",
  "title": "'Diamond Maker' Technology Simulates Subsurface Geology in Laboratories",
  "subtitle": "Allows researchers ability to speciate natural elements while investigating geologic processes",
  "category": "materials_and_coatings",
  "subcategory": "",
  "center": "MSC",
  "case_numbers": ["MSC-26347-1"],
  "patent_numbers": ["11,506,620"],
  "uspto_search_urls": [
    "https://ppubs.uspto.gov/pubwebapp/external.html?q=(11506620).pn.&db=USPAT,USOCR,US-PGPUB"
  ],
  "trl": "6",
  "abstract": "Innovators at NASA Johnson Space Center have developed a novel, double capsule control system…",
  "tech_desc": "Given the significant impact of fO2 on material properties, it is important to perform studies…",
  "benefits": [
    "Improves simulation of more extreme planetary interiors by yielding higher temperature (1600+C) and pressure values (20+GPa).",
    "Allows fO2 to be specified across wider ranges of values relevant to experimental samples."
  ],
  "applications": [
    "High pressure, high temperature geological studies",
    "Materials science and engineering"
  ],
  "tags": ["oxygen fugacity", "fO2", "astromaterial", "high pressure"],
  "images": [
    {"url": "https://technology.nasa.gov/t2media/tops/img/MSC-TOPS-89/mars_interior_2.jpg", "caption": ""}
  ],
  "pdf_url": "https://technology.nasa.gov/t2media/tops/pdf/MSC-TOPS-89.pdf",
  "detail_url": "https://technology.nasa.gov/patent/MSC-TOPS-89",
  "push_date_epoch_ms": 1640332800000,
  "contact": {
    "name": "Agency Licensing Concierge",
    "email": "[email protected]",
    "phone": ""
  }
}

Search response (multiple records)

{
  "query": "hall thruster",
  "center": "aw",
  "category": "propulsion",
  "page": 1,
  "page_size": 9,
  "results": [
    { "reference_number": "LEW-TOPS-34", "score": 17.43, "title": "Hall Effect Thruster Technologies", "...": "..." },
    { "reference_number": "LEW-TOPS-158", "score": 17.36, "title": "High Propellant Throughput Small Spacecraft Electric Propulsion Thruster", "...": "..." }
  ],
  "has_more": false
}

has_more is true whenever results.length === page_size (paginate); false whenever fewer items came back than requested. The API does not return a total count.

Direct-ID lookup, not found

{
  "success": false,
  "reason": "not_found",
  "reference_number_queried": "BOGUS-ID-123"
}

USPTO-number lookup, asked but unsupported

{
  "success": false,
  "reason": "uspto_number_search_unsupported",
  "hint": "Use ppubs.uspto.gov directly, or scan /searchosapicat with page_size=50 and filter _source.patent_number client-side."
}