Skip to main content
skill CX Codex

bigquery-patent-search — Claude Code skill: patent search

The bigquery-patent-search Claude Code skill searches patents in Google BigQuery by keywords or CPC codes. It can also retrieve a patent record by publication number.

MIT 102 installs ★ 214 updated 2026-10-06

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add RobThePCGuy/Claude-Patent-Creator --skill bigquery-patent-search -a claude-code
Open on GitHub

What bigquery-patent-search does

The skill searches for prior art, filters patents by technology area or CPC code, and retrieves records by publication number. It queries Google's public dataset of more than 100 million patents. Searches can cover titles, abstracts, and claims, with country and date filters.

It requires a Google Cloud project, the BigQuery API enabled, and configured Application Default Credentials. The authentication and project setup commands use Google Cloud SDK. Queries consume scan quota. Description retrieval is opt-in and can scan more than 1 TiB, above the default 350 GiB cap; approval is required before retrying with a higher cap.

What it helps with

  • search patents by keywords and country
  • filter documents with a CPC code
  • retrieve patent text by publication number
  • find prior art for a technology area

Example prompts for the agent

  • Find patents about blockchain authentication in the US from 2020 to 2024.
  • Search for documents under CPC code G06F16/.
  • Get the abstract and claims for patent US10123456B2.

How to install bigquery-patent-search

1Skills CLI (Vercel)

npx skills add RobThePCGuy/Claude-Patent-Creator --skill bigquery-patent-search -a claude-code

2Manually: put the files in place

~/.claude/skills/bigquery-patent-search/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/bigquery-patent-search/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about bigquery-patent-search

How do I install bigquery-patent-search in Claude Code?

Run npx skills add RobThePCGuy/Claude-Patent-Creator --skill bigquery-patent-search -a claude-code. The skill file is installed at ~/.claude/skills/bigquery-patent-search/SKILL.md.

What does bigquery-patent-search do?

It searches Google BigQuery for patents using keywords or CPC codes, and retrieves patent records and text by publication number.

What does bigquery-patent-search need to run?

It requires a Google Cloud project, the BigQuery API enabled, configured Application Default Credentials, and Google Cloud SDK for authentication and project setup.

Source text

SKILL.md · MIT
---
name: bigquery-patent-search
description: Fast, cloud-based patent searching across 100 million+ worldwide patents using Google BigQuery - keyword search, CPC classification, patent details retrieval
tools: Bash, Read, Write
model: sonnet
---

# BigQuery Patent Search Skill

> **Windows PowerShell:** replace a macOS/Linux Python invocation with `& "$HOME\patent-creator-venv\Scripts\python.exe"` and a CLI invocation with `& "$HOME\patent-creator-venv\Scripts\patent-creator.exe"`. Keep script paths quoted. Python code blocks are Python source, not shell commands.

Fast, cloud-based patent searching across 100 million+ worldwide patents using Google BigQuery.

## When to Use

Invoke this skill when users ask to:
- Search for prior art patents
- Find patents in a specific technology area
- Search by CPC classification code
- Look up patent details by publication number
- Conduct freedom-to-operate searches
- Research patent landscapes

## What This Skill Does

Provides access to Google's public patent dataset:

1. **Keyword Search** across 100M+ patents:
   - Full-text search of titles, abstracts, claims
   - Filter by country (US, EP, JP, CN, etc.)
   - Filter by filing/grant date ranges
   - Fast cloud-based queries (< 5 seconds)

2. **CPC Classification Search**:
   - Search by CPC code (e.g., "G06F16/", "H04L29/06")
   - Browse patent classifications
   - Find patents in specific technical domains

3. **Patent Details Retrieval**:
   - Get full patent text by publication number
   - Access title, abstract, claims, description
   - View CPC codes, inventors, assignees
   - See filing and grant dates

## Required Setup

This skill requires Google Cloud authentication:

**Prerequisites**:
1. Google Cloud Project (free to create)
2. BigQuery API enabled (queries consume scan quota)
3. Application Default Credentials configured

**Setup Commands**:
```bash
# Install Google Cloud SDK (if not installed)
# Visit: https://cloud.google.com/sdk/docs/install

# Authenticate
gcloud auth application-default login

# Set project (get ID from console.cloud.google.com)
"$HOME/patent-creator-venv/bin/patent-creator" config set GOOGLE_CLOUD_PROJECT your-project-id
```

Windows PowerShell:
```powershell
& "$HOME\patent-creator-venv\Scripts\patent-creator.exe" config set GOOGLE_CLOUD_PROJECT your-project-id
```

Restart Claude Code after saving the project. BigQuery is optional; the **Patent Search** skill uses SerpApi by default.

## How to Use

When this skill is invoked:

1. **Initialize BigQuery searcher**:
   ```python
   import os
   import sys
   sys.path.insert(0, os.environ.get('CLAUDE_PLUGIN_ROOT', '.'))
   from mcp_server.bigquery_search import BigQueryPatentSearch

   searcher = BigQueryPatentSearch()
   ```

2. **Search by keywords**:
   ```python
   results = searcher.search_by_keywords(
       query="blockchain authentication",
       limit=20,
       country="US",  # Optional: filter by country
       start_year=2020,  # Optional: filter by year
       end_year=2024
   )
   ```

3. **Search by CPC code**:
   ```python
   results = searcher.search_by_cpc(
       cpc_code="G06F16/",  # CPC prefix
       limit=20,
       country="US"
   )
   ```

4. **Get patent details**:
   Description retrieval is opt-in and can scan over 1 TiB, exceeding the default
   350 GiB cap. Use the free dry-run estimate and obtain approval for any cap
   increase before retrying. Omit `include_description` for a smaller lookup.
   ```python
   patent = searcher.get_patent_details(
       patent_number="US10123456B2",  # Publication number
       include_abstract=True,
       include_claims=True,
       include_description=True  # Large scan; opt in only when needed
   )
   ```

## BigQuery Dataset

Uses `patents-public-data.patents` on Google BigQuery:
- **100M+ worldwide patents**
- **12M+ US patents** with full text
- Updated weekly
- Each query consumes scan quota. A default keyword search scans hundreds of GiB.

## Search Result Format

Keyword search returns these keys (unavailable dates are null):
```python
{
    "patent_number": "US-10123456-B2",
    "application_number": "US-201615123456-A",
    "title": "Method and system for...",
    "abstract": "A system for...",
    "filing_date": "2019-01-15",
    "grant_date": "2020-06-30",
    "publication_date": "2020-06-30",
    "country": "US",
    "family_id": "12345678",
    "relevance_score": 4
}
```

Detail lookups return the identifiers, title, dates, country and family ID above,
plus `cpc_codes` and `ipc_codes`, but no `relevance_score`. They include `claims`
by default. `abstract` and `description` require their respective `include_*`
flags. Text sections are strings. Inventors, assignees and priority dates are
not returned by this implementation. Missing requested text may be empty.

## Presentation Format

Present search results as:

```
PATENT SEARCH RESULTS
====================

Query: "blockchain authentication"
Found: 247 patents (showing top 20)
Date Range: 2020-2024
Country: US

[1] US10123456B2 - System for blockchain-based authentication
    Assignee: Example Corp
    Filed: 2019-01-15 | Granted: 2020-06-30
    CPC: G06F16/245, H04L29/06

    Abstract: A system for authenticating users using blockchain
    technology with distributed ledger verification...

[2] US10234567B1 - Method of secure authentication using blockchain
    ...

---

Top 5 Most Relevant:
1. US10123456B2 (95% relevance)
2. US10234567B1 (92% relevance)
...
```

## Advanced Search Techniques

1. **Boolean Operators** in queries:
   - "blockchain AND authentication"
   - "encryption OR cryptography"
   - "(mobile OR wireless) AND security"

2. **Phrase Search** (include the double quotes in the query string):
   - `query='"distributed ledger technology"'`
   - `query='"public key infrastructure"'`

3. **CPC Code Hierarchies**:
   - "G06F" = Computing
   - "G06F16/" = Information retrieval
   - "G06F16/245" = Structured query language

## Common CPC Codes

- **G06F**: Computing, calculating, counting
- **H04L**: Digital communication
- **G06Q**: Business methods
- **H04W**: Wireless communication
- **G06N**: Computer systems based on specific models
- **G06T**: Image processing

## Error Handling

If BigQuery is not configured:
1. Check if `google-cloud-bigquery` is installed
2. Verify authentication: `gcloud auth application-default login`
3. Confirm project ID in environment: `GOOGLE_CLOUD_PROJECT`
4. Test with: `"$HOME/patent-creator-venv/bin/patent-creator" check-bigquery`

## Cost Considerations

BigQuery pricing:
- **First 1 TiB/month**: FREE
- **After 1 TiB**: $6.25 per TiB queried (on-demand)
- **Observed default keyword scan**: roughly 325-341 GiB per search, about three searches per 1 TiB allowance. These are historical measurements; use the dry-run estimate for the current query.
- Selecting fewer text fields reduces scan size. Country, date and keyword filters do not reliably reduce bytes scanned on this unpartitioned table.
- **Bytes-billed ceiling**: enforced per-query via `PATENT_BIGQUERY_MAX_BYTES_BILLED` (default 350 GiB)

## Tools Available

- **Bash**: To run Python BigQuery searches
- **Read**: To load saved search results
- **Write**: To save patent search results
- **Grep**: To search through saved results

© RobThePCGuy · MIT · RobThePCGuy/Claude-Patent-Creator

Similar skills

All skills for Data and databases →

From the same set5

Project sponsors

Companies that keep this analytics open