Patent Research Workflow

Guide for using the AI Patent Search Generator MCP connector to turn a US patent number into legal intelligence: bibliography, claims, prosecution history, PTAB validity challenges, district-court litigation, ownership chain, legal status, examiner statistics, and an AI risk verdict. All data is live from USPTO Open Data (US patents; file-wrapper coverage from 2001).

When to use this skill

Trigger when the user mentions: a US patent number (e.g. "US10867416B2", "patent 8,724,622"), "is this patent valid/alive/expired", "who owns this patent", "has this patent been challenged/litigated/sued over", "patent due diligence", "freedom to operate" (with the not-legal-advice caveat), "what are the claims of", "prosecution history / file wrapper", or "the examiner for this application".

Core workflow (patent number → intelligence)

1. Normalize the number. Tools accept loose input ("10,867,416",

"us10867416") and normalize to canonical form (US10867416B2). US patents only — say so if the user asks about EP/WO/CN numbers.

1. Start with the question's cheapest sufficient tool (don't open with the

dossier if a free lookup answers it):

• "Still in force? When does it expire?" → legal_status (free)
• "Who owns it? Chain of title?" → assignments (free)
• "Was it attacked at the PTAB? Did it survive?" → challenges (free)
• "What happened during examination?" → prosecution (free)
• "Who sued whom over it?" → litigation (2 credits)
• "How tough was the examiner?" → examiner (2 credits)

1. Go deep with `dossier` (3 credits fresh, free from 24h cache) when the

user wants the full picture: title, assignees, inventors, dates, status, independent claims, citations, family, CPC classes, similar documents. Use claims (1 credit cold, free cached) instead when they only need claim language — it's cheaper and returns every claim in full.

1. One-shot verdict with `risk_profile` (40 credits fresh, free from 24h

cache) when the user asks "how risky is this patent?" or wants due diligence: aggregates legal status + PTAB + litigation + ownership + term and returns Low/Moderate/High with a plain-English rationale and the underlying signals. Mention the cost before running it; repeat calls within 24 hours are free.

1. Interpret, don't dump.

• PTAB outcomes: "Institution Denied" / "terminated without adverse

finding" → the patent survived that challenge; a Final Written Decision finding claims unpatentable is the serious one.

• Litigation coverage is the USPTO Patent Litigation Dataset: comprehensive

2003–2016, partial to 2020, nothing after 2020. An empty result means "no suit on record in that window," not "never litigated."

• Examiner allowance rate: compare against the ~60-70% USPTO average;

pendency is reported in months from a sample of the examiner's cases.

• inForce: false with a lapse event usually means unpaid maintenance fees

— cheap to confirm via the maintenance-fee history in legal_status.

Cost & etiquette

• Free: challenges, legal_status, assignments, prosecution, balance.
• Metered (prepaid credits, 1 credit ≈ $0.10): claims 1 (cold only),

litigation 2, examiner 2, dossier 3 (fresh only), risk_profile 40 (fresh only). Cached repeats within 24h are free for dossier/claims/ risk_profile. New accounts include free starter credits.

• Check balance when a call reports insufficient credits; mention costs

before the 40-credit risk_profile run.

Output guidance

Lead with the direct answer (in force or not, survived or not, owner, risk label), then the supporting record in a compact list. Always close patent legal questions with the standard caveat: this is factual public-record reporting and machine-generated summary — not legal advice; validity, infringement, and freedom-to-operate conclusions need a registered patent attorney.