Google Patent Search

A Claude skill for searching Google Patents using plain English. Ask questions in natural language and get results from 166M+ patent publications across 17+ countries — powered by BigQuery under the hood.

What This Skill Does

This skill lets you search patent data using natural language. You ask questions in plain English, and the skill translates them into optimized BigQuery queries, runs them, and presents the results. It complements the existing USPTO Patent Search skill by adding capabilities that REST APIs can't provide:

• Full-text claims search — "find US patents granted after 2020 with claims mentioning solid-state electrolyte and lithium"
• International patents — "search European patents by Samsung about wireless power transfer"
• Large-scale analytics — "who are the top 20 filers in CPC H04L?"
• Filing trends — "show patent filing trends for wireless communication since 1990"

When Each Skill Fires

This skill (Google Patent Search) auto-triggers on:

• "find US patents with claims mentioning solid-state electrolyte and lithium"
• "search European patents by Samsung about wireless power transfer"
• "who are the top 20 filers in CPC H04L?"
• "show patent filing trends for wireless communication since 1990"

USPTO Patent Search auto-triggers on:

• "what's the status of application 16/123,456?"
• "download the file history for patent 11,887,351"
• "has patent 10,000,000 been challenged at PTAB?"
• "who owns patent 9,876,543?"
• "find patents by inventor John Smith" (defaults to USPTO)

How they complement each other

Capability	Google Patent Search	USPTO Patent Search
Claims text search	Yes (full-text)	No
International patents	17+ countries	US only
Patent analytics	Yes	Limited
Live prosecution status	No (snapshots)	Yes (live API)
PDF downloads	No	Yes
Office action details	Snapshots	Live API
Setup required	GCP + gcloud	API keys only

Rule of thumb: Use Google Patent Search for discovering and analyzing patents. Use USPTO Patent Search for specific patent lookups and prosecution data.

Data Freshness

This skill queries Google's BigQuery public patent datasets, which are periodic snapshots — not the same live index that powers patents.google.com. This means results may differ from what you see on the Google Patents website:

Dataset	BigQuery Freshness	Google Patents Web
Patent publications (titles, abstracts, claims)	Updated periodically (check Google's dataset page for latest)	Live
Assignee names	Reflects original filing only — does not include subsequent assignments	Incorporates assignment records, shows current owner
USPTO assignment records	Frozen at Feb 2017 — severely outdated	Live
PatentsView tables	Updated periodically (~quarterly)	N/A

Key implication for assignee searches: If a patent was filed by individual inventors and later assigned to a company, BigQuery will show the inventors as assignees while Google Patents web will show the company. For the most current assignee information, use the uspto-patent-search skill (which queries live USPTO APIs) or search patents.google.com directly.

Installation

Prerequisites

• Claude Code CLI installed
• Google Cloud account (free)
• Google Cloud project with BigQuery API enabled
• gcloud CLI installed

Step 1: Clone the repo

git clone https://github.com/LeonardHope/Google-Patents-Natural-Language-API-Search.git
cd Google-Patents-Natural-Language-API-Search

Step 2: Install the Claude Code skill

Symlink the entire repo into your Claude Code skills directory so all paths resolve automatically:

ln -s "$(pwd)" ~/.claude/skills/google-patent-search

This creates a symlink so Claude Code can find both the SKILL.md and the scripts directory. If you move the repo later, just re-run this command from the new location.

Step 3: Set up Google Cloud / BigQuery

# Install gcloud CLI
brew install --cask google-cloud-sdk

# Set your project
gcloud config set project YOUR_PROJECT_ID

# Authenticate
gcloud auth application-default login

Step 4: Install Python dependencies

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Step 5: Verify setup

python3 get_started.py

Cost

• Free tier: 1 TB/month of queries — no credit card or billing account needed
• Beyond the free tier: $6.25 per TB. Add a billing account to your Google Cloud project and queries automatically continue past 1 TB at this rate. There is no manual threshold to raise — billing kicks in once the free tier is exhausted.
• Built-in cost guardrails: queries over 5 GB require explicit approval, showing you the estimated cost as a percentage of the free tier before running
• Typical query costs:
  • Assignee/inventor/CPC searches: 500 MB - 2 GB
  • Keyword searches (title/abstract): 1 - 5 GB
  • Full-text claims searches: 40 - 150 GB (a single claims search can use 5-15% of the free tier)

Usage

Explicit invocation

/google-patent-search US patents with claims mentioning solid-state electrolyte and lithium
/google-patent-search European patents by Samsung about wireless power transfer
/google-patent-search top 20 filers in CPC H04L

Natural language (auto-triggers)

"find US patents granted after 2020 with claims mentioning solid-state electrolyte and lithium"
"who are the top 20 filers in CPC H04L?"
"show patent filing trends for wireless communication since 1990"

Architecture

scripts/
├── bigquery_client.py       # Auth, cost estimation, safety guardrails
├── patentsview_search.py    # Cheap patentsview.* table queries
├── publications_search.py   # Main publications table (full-text, international)
├── research_search.py       # Google AI-extracted top terms
├── prosecution_search.py    # Assignments, litigation, PTAB, PEDS, ITC
└── format_results.py        # Human-readable formatting

Every query flows through bigquery_client.py which enforces:

1. Query validation — rejects unsafe or unbounded queries
2. Dry-run cost check — estimates data scanned before executing
3. Hard cost cap — refuses expensive queries (configurable threshold)