USE THIS TOOL WHEN: You encounter errors or want to verify which APIs
are available and properly configured before starting research.

Returns status of each data source including:
- Configuration status (API keys, credentials)
- Connection availability
- Rate limit information where available

USE THIS TOOL WHEN: You need the title of a CPC code — section (G),
class (G06), subclass (G06N), main group (G06N 3/00) or subgroup
(G06N 3/08) — its place in the hierarchy, or the groups beneath it,
whether reading the codes on a patent or choosing codes for a
`.cpc.` search.

Sections and classes come from a bundled table. Subclasses and groups
come from USPTO's published CPC scheme page for the subclass (one
fetch per subclass per day, no key), so titles track the current CPC
version.

Args:
    cpc_code: Any CPC symbol form: "G06N3/08", "G06N 3/08",
           "H04B7/0417 20130101" (as PPUBS prints it), "G06N", "G06",
           "G".

Returns:
    Section: title, description and its classes. Class: section and
    class titles. Subclass or group: `title`, `level`, `hierarchy`
    from the subclass down to the code, direct `children`, and the
    source page. If the scheme page cannot be fetched, the static
    section and class titles come back with a `scheme_error`.

USE THIS TOOL WHEN: You encounter a status code in application data
and need to understand what examination stage it represents.

Args:
    code: Status code number (e.g., "30" for "Docketed New Case")

Returns:
    Status code description and examination stage.

USE THIS TOOL WHEN: You need full-text search of US patents with daily
updates, or need access to the most recent patent filings.

Args:
    query: Search query using USPTO BRS syntax. Multi-word terms are
           AND-ed across all fields by default — quote phrases that
           must appear together. Examples:
           - '"machine learning"' - exact phrase, all fields
           - machine learning - both terms anywhere (AND)
           - '"neural network".ti.' - title contains phrase
           - smith.in. AND IBM.as. - inventor Smith, assignee IBM
           - G06N3/08.cpc. - CPC classification
           Field qualifiers are dotted suffixes (.ti., .ab., .in.,
           .as., .cpc., .pn.) — the legacy slash-prefix forms
           (TTL/, IN/, AN/, CPC/) no longer work: they silently
           return 0 results, and TTL/"phrase" returns a server 500.
           Default `sort="date_publ desc"` means broad queries return
           the latest grants first — narrow with field qualifiers
           to get relevant matches.
    offset: Starting position for pagination (default: 0)
    limit: Maximum results to return (default: 20, max: 500). Results
           over the token budget are cut to 20 anyway, so raise this
           only when paging with `offset`.
    sort: Sort order (default: "date_publ desc")

Returns:
    Normalized response with patent results including GUID, title,
    abstract, dates, inventors, and classification codes.

USE THIS TOOL WHEN: You need to search pre-grant published applications
(applications publish 18 months after filing, before grant).

Args:
    query: Search query using USPTO BRS syntax (same as
           ppubs_search_patents). Multi-word terms are AND-ed —
           quote phrases that must appear together (e.g.
           '"machine learning"'). Field qualifiers are dotted
           suffixes (.ti., .ab., .in., .as., .cpc.) — the legacy
           slash-prefix forms (TTL/, IN/) no longer work.
    offset: Starting position for pagination (default: 0)
    limit: Maximum results to return (default: 20, max: 500). Results
           over the token budget are cut to 20 anyway, so raise this
           only when paging with `offset`.
    sort: Sort order (default: "date_publ desc")

Returns:
    Normalized response with application results.

USE THIS TOOL WHEN: You have a document GUID from search results
and need the patent text. A whole document runs 20-40k tokens, most
of it the description — ask for `sections` when you need less.

Args:
    guid: Document GUID (e.g., "US-9876543-B2")
    source_type: Document type - "USPAT" for patents, "US-PGPUB" for applications
    sections: Which parts to return; omit for everything. Any of
           "biblio" (front-page data: parties, dates, classification,
           references cited), "abstract", "claims", "description".
           ["biblio", "claims"] answers most questions in ~5k tokens.

Returns:
    Document with the requested sections. Empty fields and search
    highlight fields are dropped.

USE THIS TOOL WHEN: You know the patent number and need the patent
text. A whole document runs 20-40k tokens, most of it the
description — ask for `sections` when you need less.

Args:
    patent_number: Patent number. Separators, a leading "US" and a
           kind code are ignored, and design/reissue/plant prefixes
           are kept: "7123456", "US 10,000,000 B2", "D845123",
           "RE49123" all work.
    sections: Which parts to return; omit for everything. Any of
           "biblio" (front-page data: parties, dates, classification,
           references cited), "abstract", "claims", "description".
           ["biblio", "claims"] answers most questions in ~5k tokens.

Returns:
    Patent document with the requested sections. Empty fields and
    search highlight fields are dropped.

USE THIS TOOL WHEN: You need the official PDF document of a patent.
Note: Claude Desktop may not fully support PDF display.

Args:
    patent_number: Patent number ("7123456", "US 10,000,000 B2",
           "D845123" and "RE49123" all work).

Returns:
    Dictionary with base64-encoded PDF data.

USE THIS TOOL WHEN: You want to know who built on a patent, gauge its
influence, or find later art in the same space. This is the working
replacement for the decommissioned Enriched Citation API. For the
references a patent itself cites (backward citations), read its front
page instead: ppubs_get_patent_by_number(number, sections=["biblio"])
returns them in usRefGroup, foreignRefGroup and otherRefPub.

Searches the references-cited field (.urpn.) of granted patents.
Published applications do not index that field, so citing
applications that have not yet issued are not included. Verified live
2026-09-26.

Args:
    patent_number: The cited patent ("7123456", "US 10,000,000 B2",
           "D845123" and "RE49123" all work)
    offset: Starting position for pagination (default: 0)
    limit: Maximum results to return (default: 20, max: 500)
    sort: Sort order (default: "date_publ desc", newest citing patent first)

Returns:
    Normalized search response; `total` is the forward-citation count
    and each result is a citing patent with GUID, title, dates,
    inventors and classification.

USE THIS TOOL WHEN: You need prosecution/file wrapper data for an
application including status, dates, and basic metadata.

Args:
    app_num: Application number without slashes or commas (e.g., "14412875")

Returns:
    Application data including filing date, status, and basic info.

USE THIS TOOL WHEN: You need comprehensive application metadata
including examiner info, art unit, and detailed status.

Args:
    app_num: Application number without slashes (e.g., "14412875")

USE THIS TOOL WHEN: You need to understand the patent family tree -
parent applications, continuations, divisionals, and CIPs.

Args:
    app_num: Application number without slashes (e.g., "14412875")

Returns:
    Continuity data showing parent/child relationships and priority claims.

USE THIS TOOL WHEN: You need to know current and historical owners
of a patent or application.

Args:
    app_num: Application number without slashes (e.g., "14412875")

USE THIS TOOL WHEN: You need to calculate the actual expiration date
of a patent accounting for USPTO delays.

Args:
    app_num: Application number without slashes (e.g., "14412875")

Args:
    app_num: Application number without slashes (e.g., "14412875")

USE THIS TOOL WHEN: You need to find priority claims to foreign
applications that may affect the effective filing date.

Args:
    app_num: Application number without slashes (e.g., "14412875")

USE THIS TOOL WHEN: You need the complete timeline of prosecution
events including office actions, responses, and fee payments.

Args:
    app_num: Application number without slashes (e.g., "14412875")

USE THIS TOOL WHEN: You need to find office actions, responses,
notices of allowance, claims, IDS forms or examiner citations for an
application, then read one with odp_download_document.

Args:
    app_num: Application number without slashes (e.g., "14412875")
    document_code: Comma-separated USPTO document codes to keep. Common
           ones: CTNF (non-final rejection), CTFR (final rejection),
           NOA (notice of allowance), REM (applicant remarks/arguments),
           CLM (claims), SPEC (specification), IDS and 1449 (applicant
           citations), 892 (examiner citations), ADS, N417 (EFS receipt).
           `metadata.code_counts` in any response lists every code in
           the wrapper.
    direction: "INCOMING" (from applicant), "OUTGOING" (from USPTO) or
           "INTERNAL" (examiner search notes etc.)
    offset: Starting position within the filtered list (default: 0)
    limit: Max documents to return (default: 50)

Returns:
    Documents newest first, each with documentIdentifier (pass it to
    odp_download_document), documentCode, description, officialDate,
    directionCategory, page count and available formats.

USE THIS TOOL WHEN: odp_get_documents or odp_search_office_actions
gave you a document identifier and you need to read the office
action, applicant's remarks (REM), claims (CLM), notice of allowance,
petition decision or other paper itself.

Prefer format="text": it returns the paper's body as plain text (an
office action is typically 10-20k characters) from the XML USPTO
publishes with the document. Claim amendments keep their markup,
deletions as ~~text~~ and insertions as ++text++. Applicant papers
are OCR text and can carry OCR errors. Documents listed with an
"XML" format have text; others (older papers, drawings) return
NO_TEXT_AVAILABLE, and format="pdf" still works for them.

Args:
    app_num: Application number without slashes (e.g., "16123456")
    document_id: documentIdentifier from odp_get_documents, or
           document_id from odp_search_office_actions
           (e.g., "K87AK41FRXEAPX5")
    format: "text" (plain text) or "pdf" (base64 PDF, the default)
    text_offset: With format="text", the character to start from.
           Long papers come back in 24k-character pages; pass the
           previous response's next_offset to continue.

Returns:
    format="text": {"success": True, "text", "total_chars",
    "text_offset", "next_offset" (None when done), "source_file"}.
    format="pdf": {"success": True, "filename", "content_type",
    "size_bytes", "content"} with the PDF as base64. Documents over
    4 MB are refused with RESPONSE_TOO_LARGE. Requires USPTO_API_KEY.

USE THIS TOOL WHEN: You need to search applications with filtering
by applicant metadata, dates, or other criteria not available in PPUBS.

Supports both free-text queries and structured multi-field filtering.
Typed filters are AND-ed together; pass a Lucene-style string in
`query` for OR or more complex expressions.

Args:
    query: Free-text or Lucene-style query (e.g. 'neural network',
           'applicationMetaData.firstInventorName:Smith OR
           applicationMetaData.firstInventorName:Jones')
    application_number: Filter by application number (exact match)
    patent_number: Filter by patent number (exact match)
    inventor_name: Filter by inventor name (matches first inventor)
    assignee_name: Filter by applicant/assignee name (matches first applicant)
    filing_date_from: Filing date range start (YYYY-MM-DD)
    filing_date_to: Filing date range end (YYYY-MM-DD)
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Returns:
    Normalized response with matching applications.

USE THIS TOOL WHEN: You need to find office actions across many
applications by their text ("means-plus-function", "Alice",
"inherent anticipation"), by statute (101/102/103/112), by art unit
or by date, or to count them. metadata.facets gives document-code and
rejection-type counts for the whole match set, so limit=1 answers
"how many 101 rejections did art unit 3689 issue in 2026" in one call.
For one application's history, odp_get_documents is cheaper.

IMPORTANT: text_preview holds only the first 500 characters. The API
searches the full text but returns no more. To read a hit, pass its
app_num and document_id to odp_download_document(format="text").

Coverage: office actions mailed from mid-2007 on (19.9M records).
Typed filters are AND-ed; pass a Lucene string in `query` for OR.

Args:
    query: Lucene query; bare words search everything, or name a field
           such as officeActionText:"printed matter"
    app_num: Application number without slashes (e.g. "18402543")
    patent_number: Patent number; resolved to its application first
    document_code: Comma-separated codes: CTNF (non-final rejection),
           CTFR (final rejection), NOA (allowance), CTRS (restriction),
           CTAV (advisory), APEA (examiner's answer), PETDEC
    rejection_type: Comma-separated statutes, all of which must apply:
           101, 102, 103, 112, DoublePatent, Other
    art_unit: Group art unit (e.g. "3689")
    date_from: Earliest mail (submission) date, YYYY-MM-DD
    date_to: Latest mail (submission) date, YYYY-MM-DD
    record_id: officeActionRecordIdentifier from an earlier result
    sort: "newest" (default) or "oldest" by mail date
    offset: Starting position (default: 0)
    limit: Max results (default: 25, at most 100)

Returns:
    Office actions with record_id, app_num, document_code, description,
    submission_date, art_unit, title, rejection_types, document_id and
    text_preview, plus metadata.facets.

USE THIS TOOL WHEN: You need petition decisions on reviving abandoned
applications, reinstating patents after a missed maintenance fee,
delayed priority claims, petitions to make special, or supervisory
review of an examiner, or want to see how the Office of Petitions has
ruled on an argument (search its text with `query`). metadata.facets
gives outcome, deciding-office and petition-type counts for the whole
match set, so grant rates take one call.

Coverage: 1.1M decisions, 1991 on. The text is OCR and can be noisy
("EENITED STATES"), so prefer single terms or proximity
(ocrSearchText:"unintentional delay"~3) to long exact phrases. Typed
filters are AND-ed; pass a Lucene string in `query` for OR.

Args:
    query: Lucene query; bare words search everything, or name a field
           such as ocrSearchText:"unintentional delay"
    app_num: Application number without slashes
    patent_number: Patent number
    petition_type: Numeric petition type code (e.g. "533", reinstate
           patent for delayed maintenance fee) or words from its
           description (e.g. "revive")
    outcome: GRANTED, GRANTED-IN-PART, DENIED or DISMISSED
    deciding_office: e.g. "Office of Petitions", "Technology Center",
           "International Division"
    technology_center: e.g. "2600"
    date_from: Earliest decision date, YYYY-MM-DD
    date_to: Latest decision date, YYYY-MM-DD
    final_only: Only decisions marked final agency action (in the data
           so far, exactly the DENIED decisions)
    record_id: petitionDecisionRecordIdentifier from an earlier result
    include_full_text: Add each decision's full OCR text as full_text
           (about 1.5-3k characters each, prefixed by the record's
           metadata). Off by default.
    sort: "newest" (default) or "oldest" by decision date
    offset: Starting position (default: 0)
    limit: Max results (default: 25, at most 100)

Returns:
    Decisions with record_id, app_num, patent_number, dates, petition
    type, outcome, deciding office, final_decision, text (an excerpt,
    with query matches marked «like this») and documents (pass a
    documentIdentifier and app_num to odp_download_document with
    format="text" for the decision letter), plus metadata.facets.

USE THIS TOOL WHEN: You need to find bulk download datasets
available from USPTO for large-scale analysis.

Args:
    query: Search query for dataset names/descriptions
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Args:
    product_id: Dataset product identifier

USE THIS TOOL WHEN: You need to find patent validity challenges at
the Patent Trial and Appeal Board. Live via USPTO ODP v3.0.

Trial types:
- IPR: Inter Partes Review (most common, based on patents/publications)
- PGR: Post-Grant Review (broader grounds, within 9 months of grant)
- CBM: Covered Business Method (for financial method patents, sunsetted)
- DER: Derivation proceedings

Args:
    query: Full-text search query
    trial_type: Type of trial (IPR, PGR, CBM, DER)
    patent_number: Patent number being challenged
    party_name: Petitioner or patent owner name
    filing_date_from: Filing date range start (YYYY-MM-DD)
    filing_date_to: Filing date range end (YYYY-MM-DD)
    status: Proceeding status (Pending, Instituted, Terminated, FWD Entered)
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Returns:
    Normalized response with matching proceedings.

Args:
    proceeding_number: Proceeding number (e.g., "IPR2023-00001")

Returns:
    Proceeding details including parties, patent, status, and dates.

Args:
    proceeding_number: Proceeding number (e.g., "IPR2023-00001")
    document_type: Filter by type (petition, response, declaration, etc.)
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Args:
    query: Full-text search in decision text
    decision_type: Type (institution, final, termination)
    proceeding_number: Filter by proceeding number
    patent_number: Filter by patent number
    decision_date_from: Date range start (YYYY-MM-DD)
    decision_date_to: Date range end (YYYY-MM-DD)
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Args:
    decision_id: Decision identifier (trial number, e.g. IPR2022-00001)

Args:
    query: Full-text search query
    application_number: Filter by application number
    patent_number: Filter by patent number
    decision_date_from: Date range start (YYYY-MM-DD)
    decision_date_to: Date range end (YYYY-MM-DD)
    offset: Starting position (default: 0)
    limit: Max results (default: 25)

Args:
    appeal_number: Appeal number

USE THIS TOOL WHEN: You have a trademark serial number or registration
number and need its authoritative live status, mark text, owner,
international classes, filing/registration dates, or prosecution stage.
TSDR is the official USPTO status source (requires API key).

Exactly one of serial_number or registration_number is required.

Args:
    serial_number: 8-digit application serial number (e.g., "78787878")
    registration_number: Registration number (e.g., "3500027")

Returns:
    Normalized response with the trademark status record.

USE THIS TOOL WHEN: You want to see what file-wrapper documents exist
for a trademark application (office actions, responses, specimens,
registration certificates) before downloading anything. Unlike
tsdr_download_trademark_documents, this returns metadata only and is
NOT subject to the 4 req/min PDF rate limit. Use the DocumentTypeCode
values it returns (e.g. "OOA", "SPE") as the document_type filter when
downloading.

Args:
    serial_number: 8-digit application serial number (e.g., "78787878")

Returns:
    Document metadata records (type, description, dates).

USE THIS TOOL WHEN: You need office actions, responses, specimens, or
other file-wrapper documents for a trademark application.

NOTE: TSDR limits PDF downloads to 4 requests per minute per API key,
and unfiltered full wrappers can exceed 10 MB (rejected with
RESPONSE_TOO_LARGE above 4 MB). Call tsdr_list_trademark_documents
first, then filter by document_type and/or date range.

Args:
    serial_number: 8-digit application serial number (e.g., "78787878")
    document_type: Optional document type filter (e.g., "OOA" for
        outgoing office actions, "SPE" for specimens)
    date_from: Optional start date filter (YYYY-MM-DD)
    date_to: Optional end date filter (YYYY-MM-DD)

Returns:
    Dictionary with base64-encoded PDF data.

USE THIS TOOL WHEN: You need the visual representation of a design mark
or stylized word mark.

Args:
    serial_number: 8-digit application serial number (e.g., "78787878")

Returns:
    Dictionary with base64-encoded image data.

USE THIS TOOL WHEN: You need to find trademarks by mark text, owner,
goods/services description, or Nice/international class — e.g.
clearance/knockout searches for a proposed mark, or mapping a company's
trademark portfolio.

NOTE: Backed by the undocumented internal API behind tmsearch.uspto.gov
(the TESS replacement) — same risk profile as PPUBS; may break without
notice. For authoritative status of a specific mark, use
tsdr_get_trademark_status. Federal marks only (no common-law/state marks).

Args:
    query: Raw query string for advanced searches
    mark_text: Word mark text to search (e.g., "ACME")
    owner_name: Owner/applicant name to search
    goods_services: Goods/services description terms (e.g., "athletic
        footwear") — useful for clearance searches across wording
    registration_number: Exact registration number to look up
    international_class: Nice class number 1-45 (e.g., "9" for software;
        use get_trademark_class_info to look up classes)
    status_filter: "live" (active marks), "dead" (abandoned/cancelled/
        expired), or omit for both
    offset: Starting position for pagination (default: 0)
    limit: Maximum results to return (default: 25, max: 100)

Returns:
    Normalized response with matching trademark records.

USE THIS TOOL WHEN: You have a serial number from tm_search_trademarks
results and want the full indexed record. For authoritative live status,
prefer tsdr_get_trademark_status.

Args:
    serial_number: 8-digit application serial number (e.g., "78787878")

Returns:
    Normalized response with the trademark record.

USE THIS TOOL WHEN: You need to trace ownership history of a trademark,
find the current owner, or find marks transferred to/from a company.
Covers recorded assignments from 1955 to present, via the USPTO
Assignment Center public API (no API key required).

At least one filter is required; multiple filters combine (AND).

Args:
    serial_number: Trademark application serial number
    registration_number: Trademark registration number
    assignee_name: Assignee (new owner) name
    assignor_name: Assignor (previous owner) name
    reel_frame: Recordation reel/frame identifier (e.g., "9006/0093")
    offset: Starting position for pagination (default: 0)
    limit: Maximum results to return (default: 25)

Returns:
    Normalized response with assignment records (reel/frame, assignors,
    assignees, conveyance, and affected properties).

USE THIS TOOL WHEN: You need to know what goods (classes 1-34) or
services (classes 35-45) a trademark class covers, or pick the right
classes for a clearance search.

Args:
    class_number: International class number (e.g., "9" for computer
        software, "25" for clothing, "42" for software services)

Returns:
    Class title, type (goods/services), and description.

USE THIS TOOL WHEN: Trademark data contains a numeric status code you
need to interpret (e.g., "700" = Registered, "686" = Published for
opposition).

Args:
    code: Trademark status code (e.g., "700")

Returns:
    Status code description and prosecution stage.

USE THIS TOOL WHEN: You need which claims an examiner rejected, objected
to or allowed in each office action, and under which statute (101, 102,
103, 112, double patenting); or claim-level 101 analytics across
applications, such as actions citing Alice or Mayo in an art unit
(case_law="alice", art_unit="3689"). For the action's text, search with
odp_search_office_actions and read it with odp_download_document.

Results come grouped by office action. statutes and case_law describe
the whole action; claims sums up the claim numbers by what happened to
them; claim_groups keeps the dataset's rows (each with action,
legal_section, form paragraph and claims). The rows are extracted by
machine from the action's text, so treat them as a guide and read the
action for anything that matters.

Coverage: actions mailed from 2012 on (93M rows), appearing about a
week after mailing. Pagination counts rows, not actions; an action can
straddle two pages. Continue from metadata.next_offset, not offset +
count: when the token budget drops actions, next_offset points at the
first row of the first one dropped.

Args:
    application_number: Application number without slashes (e.g. "18402543")
    mail_date: Mail date of one action, YYYY-MM-DD
    patent_number: Patent number; resolved to its application first
    document_id: File-wrapper documentIdentifier of one action
    statute: Comma-separated statutes, all of which the action must
        apply: 101, 102, 103, 112, DP (double patenting)
    case_law: Comma-separated cases, all of which the action must cite:
        alice, mayo, myriad, bilski
    document_code: Comma-separated codes, e.g. CTNF (non-final), CTFR (final)
    art_unit: Group art unit (e.g. "3689")
    date_from: Earliest mail date, YYYY-MM-DD
    date_to: Latest mail date, YYYY-MM-DD
    query: Extra Solr clause, AND-ed with the filters
        (e.g. nationalClass:"705")
    offset: Row offset (default: 0)
    limit: Rows to read (default: 50, at most 200)

Returns:
    Office actions with document_id, app_num, document_code, mail_date,
    art_unit, statutes, case_law, claims and claim_groups. total counts
    rows; metadata.next_offset continues the scan.

USE THIS TOOL WHEN: You need the references an examiner cited against
an application, and under which statute (102 or 103); or, in reverse,
which applications' office actions cited a given patent or publication
(cited_reference="US 2013/0120945 A1"). For citations on an issued
patent's front page, ppubs_get_patent_by_number(sections=["biblio"])
is cheaper; for later patents citing it, ppubs_get_citing_patents.

Coverage: office actions from about 2012 on; rows appear about a week
after mailing. Rows carry no mail date, so mail_date (with an
application or patent number) finds that day's actions in the
rejection dataset first.

Args:
    application_number: Application number without slashes (e.g. "18402543")
    mail_date: Mail date of one action, YYYY-MM-DD (needs
        application_number or patent_number)
    patent_number: Patent number; resolved to its application first
    document_id: File-wrapper documentIdentifier of one action
    cited_reference: A cited US patent or publication number, with or
        without "US", separators or kind code
    legal_section: Comma-separated sections the reference was applied
        under: 102, 103
    examiner_cited: True for references the examiner cited
    art_unit: Group art unit (e.g. "1798")
    tech_center: Technology center (e.g. "1700")
    query: Extra Solr clause, AND-ed with the filters
    offset: Row offset (default: 0)
    limit: Rows to read (default: 50, at most 200)

Returns:
    Citations with app_num, document_id, reference (as written),
    reference_number (digits, when USPTO parsed them), legal_section,
    action, examiner_cited, applicant_cited, relied_on_in_action,
    art_unit and tech_center.