# smythmyke/patent-search-mcp-server [Health: Active]

**Category:** ⚖️ Legal  
**Repository:** https://github.com/smythmyke/patent-search-mcp-server  
**GitHub Stars:** 2  
**npm Downloads (last month):** 295  
**Views:** 6  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/smythmyke-patent-search-mcp-server

## Description
Patent intelligence and prior-art research for the AI Patent Search Generator. Eleven tools: full patent dossier (bibliography, claims, citations, family, classifications, examiner stats); USPTO prosecution-history file wrappers; AI Office Action analysis (rejection grounds, cited prior art, suggested response arguments); Boolean query generator; multi-strategy patent search (telescoping / onion-ring / faceted); similar-document ranking; citation graph (backward + forward, examiner-cited flagged); family lookup; and CPC classification lookup. Install: npx -y patent-search-mcp-server.

## Tools
Capabilities this server exposes over MCP:

- **balance** — Return current credit balance + subscription status. Free. No arguments.
- **dossier** — Full patent intelligence: bibliographic, claims, citations, family, classifications, similar documents, examiner stats — all bundled. **3 credits on fresh fetch; free on 24h cache hit.**
- **prosecution** — USPTO file-wrapper documents (Office Actions, responses, amendments, etc.) for a US patent. Free.
- **oa_analyze** — AI analysis of a USPTO Office Action — rejection grounds, cited prior art, suggested response arguments. **First 5 analyses per application are free; subsequent analyses cost 1 credit each.**
- **examiner** — Examiner name, art unit, total applications, allowance rate, average pendency. Free.
- **query** — Single optimized Boolean query string for **manual paste** into Google Patents. Does NOT execute. **1 credit.**
- **search** — **Executes** a multi-query patent search against Google Patents server-side and returns ranked, deduplicated hits. **1 credit.**
- **similar** — Google Patents' similar-documents ranking for a given patent. Free.
- **citations** — Backward + forward citations for a patent. Each citation flags whether it was examiner-cited. Free.
- **family** — Patent family — continuations, divisionals, foreign counterparts. Free.
- **cpc** — CPC classification code lookup. Free. v1.0 covers all sections + ~80 common subclasses; subgroup descriptions land in v1.2.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "patent-search-mcp-server": {
    "command": "npx",
    "args": ["-y","patent-search-mcp-server"],
    "env": {
      "PATENT_SEARCH_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `PATENT_SEARCH_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What smythmyke/patent-search-mcp-server MCP server does

The smythmyke/patent-search-mcp-server MCP server exposes patent research functions through the Model Context Protocol. It connects to the AI Patent Search Generator service, which provides access to USPTO public data and Google Patents search capabilities.

Its research tools can assemble a patent dossier containing bibliographic data, claims, citations, family relationships, classifications, similar documents, and examiner statistics. Separate tools retrieve claims, prosecution-history documents, examiner information, patent families, citation networks, similar-document rankings, and CPC classifications.

For prior-art work, the server can create a Boolean query for manual use in Google Patents or execute a multi-query search itself. Search strategies include telescoping, onion-ring, and faceted approaches. Results are ranked and deduplicated. Other supported workflows include AI analysis of Office Actions, claim-element charts, CPC suggestions, patent legal status, assignments, term data, PTAB challenges, district-court litigation, and company litigation lookups where those tools are available through the configured service.

## How it works

An MCP client starts the server with Node.js and `npx`, then sends tool calls over the standard MCP connection. The server forwards authenticated requests to the configured API. Patent identifiers can be supplied as patent numbers; some prosecution requests also accept a US application number.

The `dossier` tool combines several research categories in one response. `prosecution` retrieves file-wrapper materials such as Office Actions, responses, and amendments. `oa_analyze` can select the latest Office Action automatically or analyze a document identified by an application number and document ID. `query` returns text for manual pasting and does not run a search, while `search` runs the multi-query search on the server.

Credit use depends on the operation and cache state. Dossier retrieval costs three credits on a fresh request but is free for a 24-hour cache hit. Search and query each cost one credit. The first five Office Action analyses for an application are free, and later analyses cost one credit. Claims and claim charts can also be free when dossier data is cached, with higher charges for cold requests. The balance tool reports credits and subscription status.

## Setup and configuration

The smythmyke/patent-search-mcp-server MCP server requires an API key created in the AI Patent Search Generator Chrome extension's Admin tab. Node.js 18 or later is needed for a local installation; an `npx` configuration does not require a separate local package installation.

A typical MCP configuration runs `npx -y patent-search-mcp-server` and passes `PATENT_SEARCH_API_KEY` in the environment. Claude Code uses `~/.claude/mcp.json` or a project `.mcp.json`; Claude Desktop uses its platform-specific configuration file; Cursor can add the same server configuration through Settings → MCP. The optional `PATENT_SEARCH_API_BASE` variable overrides the default API endpoint.

For local development, clone the repository, install dependencies with npm, build the project, and point the client at `dist/index.js` with Node. API keys should not be committed to source control; leaked keys can be revoked from the extension's Admin tab.

## Tools and capabilities

- Retrieve bundled patent intelligence with `dossier`.
- Fetch claims, claim charts, prosecution records, examiner statistics, assignments, legal status, and term information.
- Analyze Office Action rejection grounds, cited prior art, and possible response arguments.
- Generate Boolean queries or execute ranked, deduplicated patent searches.
- Inspect similar documents, backward and forward citations, examiner-cited references, families, and CPC codes.
- Check credit balance and subscription status.

## Limitations and notes

The smythmyke/patent-search-mcp-server MCP server depends on a valid API key and available credits for metered operations. Google Patents search is executed only by `search`; `query` produces a string for manual use. CPC coverage in v1.0 includes all sections and about 80 common subclasses, while subgroup descriptions are planned for v1.2. The curated CPC suggestion dataset may miss niche chemistry and biotechnology topics. Rate-limit, missing-key, and exhausted-credit errors require configuration changes, waiting, or additional credits.

_Full upstream README: https://allmcps.com/mcp/smythmyke-patent-search-mcp-server/readme_