patentsview-mcp

12 million US patents, searchable in seconds — prior art and IP intelligence for AI.

---

What Is This?

The United States Patent and Trademark Office has issued over 12 million patents since 1976 — the most comprehensive record of technological innovation in existence. PatentsView, a USPTO initiative, makes this data available through a structured API. But the API requires expertise to query, the data model is non-trivial, and it's completely inaccessible to AI assistants natively.

patentsview-mcp is a Model Context Protocol actor that turns 12 million patents into a conversational interface. Ask Claude to find prior art, map a competitor's IP portfolio, identify the most prolific inventors in a technology area, or analyze patent trends across CPC codes — and get structured results back in seconds.

The actor includes graceful API fallback — when the PatentsView API is temporarily unreachable (planned maintenance windows are common), the actor transparently serves from a cached snapshot and notifies the caller, ensuring your workflows never break unexpectedly.

This is the tool that IP attorneys used to pay $50,000/year for from specialized patent analytics vendors. Now it's a $49/month MCP tool call away.

---

Who Uses This?

User Type	What They Do Today	What This Enables
IP Attorneys	Manual USPTO full-text search, expensive patent analytics platforms	Conversational prior art search, portfolio mapping in minutes
R&D Directors	Quarterly patent landscape reports from consultants	Self-serve technology area mapping, inventor network analysis
Startup Founders	Expensive FTO (Freedom to Operate) searches	First-pass FTO risk assessment before engaging outside counsel
M&A Analysts	IP due diligence via manual patent review	Automated portfolio assessment, claim coverage analysis
University Tech Transfer	Finding commercialization partners for licensed IP	Identify companies already patenting adjacent to the university's technology
Patent Monetization Funds	Identifying undervalued patent portfolios	Systematic portfolio screening across thousands of assignees
Competitive Intelligence Teams	Tracking competitor R&D via patent filings	Real-time assignee monitoring, CPC code trend analysis
Academic Researchers	Studying innovation patterns and inventor networks	Citation analysis, geographic diffusion, temporal trends

---

Feature Overview

Feature	Detail
Full-Text Keyword Search	Search across title, abstract, and claims text
Assignee Search	Find all patents held by a company or individual
CPC Code Filter	Search by Cooperative Patent Classification codes
Inventor Search	Find patents by named inventor, with disambiguation
Date Range	Application date and grant date filtering
Citation Analysis	Forward and backward citation counts
Geographic Filter	Filter by inventor or assignee country/state
Portfolio Analysis	Aggregate stats for an assignee's full portfolio
Technology Trend Data	Patent volume by CPC code over time
Co-inventor Networks	Map inventor collaboration networks
Claim Count	Independent and dependent claim counts
Graceful API Fallback	Cached snapshot served when live API is unreachable
MCP + Batch	Real-time tool calls and bulk processing both supported

---

Quick Start

MCP Mode

{
  "mcpServers": {
    "patentsview": {
      "command": "npx",
      "args": [
        "-y",
        "@apify/mcp-client",
        "--actor-id",
        "your-username/patentsview-mcp",
        "--token",
        "YOUR_APIFY_TOKEN"
      ]
    }
  }
}

Ask Claude:

"Search patents for solid-state battery technology filed since 2019. Show me the top 10 assignees by patent count, and flag any that aren't the major auto manufacturers or battery companies — those are the ones I want to investigate."

Batch Mode — Portfolio Screen

{
  "mode": "batch",
  "portfolioScreening": {
    "assignees": ["Apple Inc.", "Google LLC", "Microsoft Corporation", "Amazon Technologies Inc."],
    "cpcCode": "G06F40",
    "dateFrom": "2020-01-01",
    "metrics": ["count", "citationImpact", "claimBreadth"]
  }
}

API Mode

curl -X POST https://api.apify.com/v2/acts/your-username~patentsview-mcp/run-sync \
  -H "Authorization: Bearer YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "tool": "search_patents",
    "keyword": "transformer neural network attention mechanism",
    "dateFrom": "2017-01-01",
    "maxResults": 100,
    "sortBy": "citationCount"
  }'

---

MCP Tools Reference

search_patents

Search 12M+ US patents with flexible filtering.

Parameters:

Parameter	Type	Required	Description
keyword	string	No	Full-text search in title, abstract, or claims
searchIn	array	No	Fields to search: title, abstract, claims (default: all)
assigneeName	string	No	Assignee/company name (partial match)
inventorName	string	No	Inventor name (disambiguation applied)
cpcCode	string	No	CPC classification code (e.g., G06F40, A61K31)
applicationDateFrom	string	No	Application filed date start (YYYY-MM-DD)
applicationDateTo	string	No	Application filed date end (YYYY-MM-DD)
grantDateFrom	string	No	Grant date start (YYYY-MM-DD)
grantDateTo	string	No	Grant date end (YYYY-MM-DD)
country	string	No	Inventor or assignee country (ISO 2-letter)
state	string	No	US state (inventor location)
minCitations	integer	No	Minimum forward citation count
sortBy	string	No	relevance, grantDate, applicationDate, citationCount
maxResults	integer	No	Max results (default: 50, max: 1000)
includeFields	array	No	Specific fields to return (default: all)

Response:

{
  "patents": [
    {
      "patentNumber": "US11234567B2",
      "title": "Attention mechanisms for neural machine translation",
      "abstract": "A method and system for...",
      "grantDate": "2022-03-15",
      "applicationDate": "2019-11-08",
      "assignees": [
        { "name": "Google LLC", "country": "US", "state": "CA" }
      ],
      "inventors": [
        { "name": "Ashish Vaswani", "disambiguatedId": "inv-12345" }
      ],
      "cpcCodes": ["G06N3/0454", "G06F40/30"],
      "claims": {
        "totalCount": 20,
        "independentCount": 3
      },
      "citations": {
        "forward": 847,
        "backward": 34
      },
      "applicationNumber": "US16679234"
    }
  ],
  "totalCount": 4821,
  "apiStatus": "live",
  "queryMetadata": {
    "executedAt": "2024-01-15T14:23:11Z",
    "dataSource": "PatentsView API v1",
    "fallbackUsed": false
  }
}

get_assignee_portfolio

Retrieve comprehensive portfolio statistics for a company or individual assignee.

Parameters:

Parameter	Type	Required	Description
assigneeName	string	Yes	Company or individual name
dateFrom	string	No	Portfolio date filter start
dateTo	string	No	Portfolio date filter end
breakdown	string	No	cpc, year, inventor, geography
includeExpired	boolean	No	Include expired patents (default: true)

Response includes:

• Total patent count (by year)
• CPC code distribution (technology areas)
• Top inventors by patent count
• Citation impact metrics (average forward citations)
• Geographic coverage
• Claim breadth statistics
• Estimated portfolio age and maintenance status

get_patent_stats

Aggregate analytics for a technology area or search query.

Parameters:

Parameter	Type	Required	Description
keyword	string	No	Technology area keyword
cpcCode	string	No	CPC code to analyze
dateFrom	string	No	Start of analysis period
dateTo	string	No	End of analysis period
metrics	array	No	trendByYear, topAssignees, topInventors, geographicDistribution, citationImpact

---

API Fallback System

PatentsView schedules maintenance windows (typically announced on their status page). This actor implements a three-tier fallback strategy:

1. Live API — All queries go to api.patentsview.org by default
2. Cached Snapshot — If the live API returns errors or times out, the actor falls back to a locally cached dataset snapshot (updated monthly)
3. Partial Results — For queries that partially complete before a timeout, partial results are returned with a partialResultsWarning flag

The apiStatus field in every response will be "live", "cached_snapshot", or "partial" so your workflow always knows what it received.

{
  "apiStatus": "cached_snapshot",
  "cacheDate": "2024-01-01",
  "warning": "PatentsView API unreachable. Serving from January 2024 snapshot. Data may be up to 30 days stale."
}

---

Use Case Recipes

Recipe 1: IP Attorney — Prior Art Search

Scenario: Your client has invented a new method for privacy-preserving federated machine learning. Before filing, you need first-pass prior art.

Prompt:

"Search patents for 'federated learning' AND 'differential privacy' filed between 2015 and 2023. Sort by forward citations (most cited first). Show me the top 20 results with patent numbers, titles, assignees, filing dates, and claim counts. Flag any with more than 100 forward citations — those are the references I need to address in my prosecution strategy."

Recipe 2: Startup Founder — FTO Risk Assessment

Scenario: You're building a product that uses on-device neural networks for real-time audio processing. You need a first-pass FTO screen.

Prompt:

"Search for patents covering on-device neural network inference for audio processing. Focus on patents granted in the last 5 years that are still active. Who are the top 5 assignees? Do any of the independent claims broadly cover 'neural network inference on edge devices for audio'? Give me the patent numbers of the highest-risk ones to flag for my patent counsel."

Recipe 3: M&A Due Diligence — IP Portfolio Assessment

Prompt:

"Pull the complete patent portfolio for 'Synapse AI Inc.' Break it down by: total patents, year of filing trend (growing or declining?), technology area by CPC code, average forward citation count versus industry average, and any patents with zero forward citations filed before 2018 (likely abandoned in practice). I need this to assess whether their IP is real or paper."

Recipe 4: R&D Director — Competitive Landscape

Prompt:

"I need a competitive patent landscape for solid-state batteries. Search CPC code H01M10/0562. For the period 2018–2024: who are the top 10 assignees by patent count? How does Toyota compare to Samsung, CATL, and QuantumScape? Show me the trend by year for each — I want to know who's accelerating their filing pace."

Recipe 5: University Tech Transfer — Commercialization Partner Discovery

Prompt:

"Our university has licensed a patent on photonic computing for matrix multiplication (US10987654). Find all companies that have filed patents citing this patent forward, OR that are filing in CPC codes H04B10 and G06F7 since 2020. These are potential licensees or acquirers. Rank by patent count in adjacent areas."

Recipe 6: Patent Monetization — Portfolio Screening

Prompt:

"Search for patents assigned to companies with fewer than 10 employees (use state='DE' as proxy for holding companies) in CPC code G06Q20 (payment systems) filed 2010–2015 with more than 50 forward citations. These are the needle-in-haystack high-value underutilized patents I need to evaluate."

---

Claude Desktop Configuration

{
  "mcpServers": {
    "patentsview": {
      "command": "npx",
      "args": [
        "-y",
        "@apify/mcp-client",
        "--actor-id",
        "your-username/patentsview-mcp",
        "--token",
        "apify_api_XXXXXXXXXXXXXXXXXXXX"
      ]
    }
  }
}

Verification prompt: "How many patents has Tesla filed in battery technology since 2020?"

---

Input Schema (Full)

{
  "tool": {
    "type": "string",
    "enum": ["search_patents", "get_assignee_portfolio", "get_patent_stats"],
    "description": "Which tool to invoke"
  },
  "keyword": { "type": "string" },
  "assigneeName": { "type": "string" },
  "inventorName": { "type": "string" },
  "cpcCode": {
    "type": "string",
    "description": "CPC classification code. Can be section (A-H), class (A01), subclass (A01B), or full code (A01B1/00)"
  },
  "applicationDateFrom": { "type": "string", "format": "date" },
  "applicationDateTo": { "type": "string", "format": "date" },
  "grantDateFrom": { "type": "string", "format": "date" },
  "grantDateTo": { "type": "string", "format": "date" },
  "minCitations": { "type": "integer", "minimum": 0 },
  "sortBy": {
    "type": "string",
    "enum": ["relevance", "grantDate", "applicationDate", "citationCount"],
    "default": "relevance"
  },
  "maxResults": {
    "type": "integer",
    "minimum": 1,
    "maximum": 1000,
    "default": 50
  },
  "breakdown": {
    "type": "string",
    "enum": ["cpc", "year", "inventor", "geography"]
  },
  "metrics": {
    "type": "array",
    "items": {
      "type": "string",
      "enum": ["trendByYear", "topAssignees", "topInventors", "geographicDistribution", "citationImpact"]
    }
  }
}

---

Pricing & Usage

Tier	Price	Includes
Free	$0	50 searches/month
Starter	$29/month	2,000 searches/month
Professional	$99/month	20,000 searches/month + portfolio analysis
Enterprise	$399/month	Unlimited + batch + priority queue

Patent analytics platforms (Derwent Innovation, PatSnap, Innography) charge $15,000–$80,000/year for similar capabilities. This actor delivers 80% of the research value at 1% of the cost.

---

Data Source & Freshness

• Primary source: PatentsView API (api.patentsview.org) — USPTO-funded research initiative
• Coverage: US utility patents granted from 1976 to present
• Update frequency: PatentsView updates monthly with USPTO bulk data
• Inventor disambiguation: PatentsView applies ML-based inventor disambiguation (same inventor, multiple name spellings resolved)
• CPC codes: Full CPC classification as assigned by USPTO examiners
• Fallback snapshot: Refreshed monthly; age displayed in cacheDate field

---

Technical Notes

• PatentsView API enforces a 500ms rate limit between requests; this actor manages queueing automatically
• Keyword search uses PatentsView's native full-text index — not regex; use natural language terms
• CPC code searches support wildcards at the subgroup level (e.g., G06F40/* returns all G06F40 subgroups)
• Inventor disambiguation is PatentsView's own — approximately 95% accurate; exact inventor queries should validate against the disambiguatedId field
• The actor does not parse or interpret individual patent claims — it returns claim text and counts as-is
• PatentsView does not include design patents, plant patents, or PCT applications (US national phase entries only)
• Results are limited to granted patents; pending applications require USPTO PAIR API (separate actor)

---

Support & Issues

• GitHub: github.com/your-username/patentsview-mcp
• PatentsView Status: status.patentsview.org
• Email: support@your-domain.com

Data sourced from PatentsView (patentsview.org), a USPTO-funded research project. This actor is not affiliated with the USPTO or PatentsView.