Patent Maintenance Fees MCP Server

@pipeworx/patent-maintenance-fees

Tell whether a US patent is still in force, or has lapsed for non-payment of maintenance fees — the one question the patents pack cannot answer, because its prosecution-event data has no fee field at all.

Part of Pipeworx — an MCP gateway connecting AI agents to 1831+ live data sources. This is an independent, unofficial integration — not affiliated with, endorsed by, or published by the upstream provider.

Tools

• get_patent_maintenance_status(patent_number) — every recorded maintenance-fee event for one patent (payments, surcharges, expiration for non-payment, reinstatement, entity-status changes), a computed status, and a 4th/8th/12th-year fee-window breakdown (due date, grace-period end, whether paid). Design (D-prefixed) and plant (PP-prefixed) patents are statutorily exempt from maintenance fees and report subject_to_maintenance_fees:false instead of a bare no-match.
• search_maintenance_fee_events(event_code?, patent_number_prefix?, date_from?, date_to?, limit?) — an exact total_matching count first, then a capped, newest-first sample of the matching events. Useful for questions like "how many patents expired for non-payment last quarter" — pass event_code: "EXP." and a date range.

Auth

Keyless.

Data source

USPTO, Patent Maintenance Fee Events — the Open Data Portal's bulk-data product recording every maintenance-fee event for US patents granted 1981-09-01 to present. US federal data, public domain.

Why this is a local-copy pack, not a live proxy

Checked first, per CLAUDE.md's local-copy rule: the ODP's Patent File Wrapper search/get-by-number API — the one mcps/patents already calls — returns prosecution events (eventDataBag: filing, publication, grant, office actions) but has no maintenance-fee field at all. Maintenance-fee history is published only as the "PTMNFEE2" bulk product, a single weekly-cumulative file, with no per-patent query endpoint. The legacy keyless host this product used to publish from (bulkdata.uspto.gov) no longer resolves in DNS as of 2026-10-17, confirmed against two independent public resolvers (NXDOMAIN/NODATA, not a transient block) — the product is reachable today only through ODP's own authenticated bulk-data API, which answers every /api/v1/* path 401/403 without an API key (the same credential type as the patents pack's PLATFORM_USPTO_KEY). A bulk file with no per-record API is exactly the case CLAUDE.md's sourcing rule calls for a refreshed backing store rather than a per-call upstream fetch.

This pack reads a Supabase mirror (uspto_maintenance_fee_events, schema in supabase/migrations/260_uspto_maintenance_fee_events.sql), loaded by scripts/ingest-uspto-maintenance-fees.mjs (.github/workflows/uspto-maintenance-fees-refresh.yml, dispatch-only — see below). Every response carries data_as_of from the mirror's own loaded_at.

Not yet provisioned (as of the commit that added this pack)

Two things still need to happen before uspto_maintenance_fee_events has any rows, and both are outside what a commit-only build session can do:

1. An ODP API key as the USPTO_ODP_API_KEY repo secret. Get one free at https://data.uspto.gov/myodp (same credential type the gateway already holds server-side as PLATFORM_USPTO_KEY for the patents pack — stored again here because the refresh job runs outside the gateway Worker).
2. The first dispatch of uspto-maintenance-fees-refresh.yml. The ingest script's fetch/discovery step (finding and downloading the latest MaintFeeEvents_*.zip via the ODP bulk-data API) follows the documented shape of that API but has never run against the live endpoint — no key was available in the build session that wrote it. It fails loudly and names the actual response shape if that shape doesn't match, rather than mis-parsing silently; see the script's header. The fixed-width record layout it parses afterward (57 bytes/record, USPTO's own documented column offsets) is independently corroborated and covered by unit tests in scripts/__tests__/ingest-uspto-maintenance-fees.test.mjs.

Until both land, every tool call fails loudly with a dataset_unavailable error rather than silently reporting every patent as having no fee events — see "Empty-mirror guard" below.

Scale and load posture

The full file is ~27M event rows (one row per event, not per patent) as of 2026-10. The refresh workflow does a full TRUNCATE + psql \copy reload every run — the file is cumulative, so reprocessing it in full reproduces the correct complete history every time, the same "no delta logic to get subtly wrong" posture as scripts/ingest-faa-aircraft-registry.mjs. Because a 27M-row reload is a big load, the workflow is workflow_dispatch only (same shape as sec-13f-load.yml), not a cron — the first backfill should be run as a monitored one-off with a Supabase latency/PGRST002 check before and after, per Bruce's 2026-09-30 "supa stagger" ruling for SEC 13F.

Empty-mirror guard

Same silent-zero class fixed for faa_reg_aircraft (fleet #2790): an empty backing table still answers REST queries successfully with zero rows, so a naive read would report every patent as having no maintenance-fee events — indistinguishable from a genuine miss against good data. Both tools check for at least one loaded row first and throw a loud dataset_unavailable error naming the real condition instead.

Scope — what this pack does NOT determine

Maintenance-fee lapse is one way a patent can stop being enforceable, not the only one. This pack does not determine:

• Full-term (20-years-from-filing) expiration. A patent with a perfect fee-payment record still expires at the end of its statutory term.
• Invalidity, reexamination outcomes, or litigation status.
• Design and plant patents at all — see below.

Pair with the patents pack for prosecution history and the epo-ops pack for non-US patents.

Design and plant patent exemption

35 U.S.C. 41(b) applies maintenance fees to utility and reissue patents only. Design (D-prefixed) and plant (PP-prefixed) patent numbers never generate a maintenance-fee event and never appear in this dataset. get_patent_maintenance_status detects the prefix and reports subject_to_maintenance_fees:false with an explanation, rather than a bare "no events found" that would read identically to a real gap.

Leading zeros — an open normalization question

Whether the file's fixed-width patent-number column keeps or strips leading zeros has not been verified against a live file sample (no API key was available while building this pack — see "Not yet provisioned" above). get_patent_maintenance_status queries both the normalized form and the same value with leading zeros stripped, rather than assuming one.

Event codes

USPTO's own long-stable maintenance-fee event codes (payments by window/entity-size/public-law era, surcharges, EXP./EXPX expiration/reinstatement, entity-status changes, petitions) are decoded in src/index.ts's EVENT_CODES table with the raw code always included alongside the description — an uncommon or future code not in the table comes back with the raw code and a null description rather than a guess.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "patent-maintenance-fees": {
      "url": "https://gateway.pipeworx.io/patent-maintenance-fees/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/patent-maintenance-fees/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — ask_pipeworx, discover_tools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

This is deliberate, not multiplexing by accident. The meta-tools are what let a scoped connection answer a question this pack does not cover — via ask_pipeworx, which routes across the whole catalog — without you adding a second MCP server. There is currently no way to mount a pack endpoint without them; if the extra schemas cost you more context than the routing is worth, connect to the full gateway once rather than to several pack endpoints.

Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:

{
  "mcpServers": {
    "pipeworx": {
      "url": "https://gateway.pipeworx.io/mcp"
    }
  }
}

Both URLs reach the same gateway and the same 1831+ data sources. The only difference is which pack's tools are listed directly; ask_pipeworx reaches all of them from either one.

No MCP client? Call it over HTTP

curl -X POST https://gateway.pipeworx.io/v1/tools/get_patent_maintenance_status \
  -H 'Content-Type: application/json' \
  -d '{"patent_number":"10835610"}'

No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/get_patent_maintenance_status. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.

Standalone (no gateway account)

This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:

{
  "mcpServers": {
    "patent-maintenance-fees": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-patent-maintenance-fees"]
    }
  }
}

Or run it directly to confirm it starts:

npx -y @pipeworx/mcp-patent-maintenance-fees

It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call for only this pack's tools — none of the shared meta-tools the gateway connection above adds. Same source, same tools, no ask_pipeworx routing.

Using with ask_pipeworx

Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:

ask_pipeworx({ question: "your question about Patent Maintenance Fees data" })

The gateway picks the right tool and fills the arguments automatically.

More

• Docs and guides
• pipeworx.io

License

MIT