Sandbox
@patsnap/patent-literature-search-mcp

MCP server for patent and literature search

Patent & Literature Search MCP connects an agent to Patsnap’s patent and academic literature databases. It supports natural-language search across patents and papers, plus semantic, keyword, BM25, and structured queries, then returns selected results in AI-readable Markdown. It works as a hosted MCP endpoint or as a local Docker bridge that forwards requests to Patsnap when `PATSNAP_API_KEY` is set.

79 stars3 forksJavaScriptUpdated 1mo ago
Who it's for

Builders who want Claude Code, Cursor, or another MCP client to search patents and scientific literature from one place.

What it delivers

You can ask your agent to research prior art and scientific literature without manually switching between search sites.

What it does

Patent search

Search patents with filters for assignee, inventor, IPC class, legal status, jurisdiction, dates, and citations.

Scientific literature search

Search papers and preprints with semantic, keyword, BM25, and structured search.

Cross-domain research

Run patent and paper searches in one request and combine the results for agent reasoning.

Record retrieval

Fetch selected patent publications or literature records as Markdown, including claims, citations, legal data, and family data where available.

Hosted and local access

Use the hosted MCP endpoint directly, or run the repository’s Docker bridge with an API key.

How to get it

  1. 1Run
    claude mcp add --transport http search-tool \
      "https://connect.patsnap.com/2b0355/logic-mcp?apikey=YOUR_API_KEY"

README

Patent & Literature Search MCP

Smithery badge

Search global patents and scientific literature in natural language, directly inside Claude, Cursor, or any MCP-compatible AI tool. The MCP supports semantic, keyword, BM25, and structured search, then retrieves selected records as AI-readable Markdown.

Start free: Register on Patsnap Open Platform to receive 10,000 free credits. No credit card is required.

About Patsnap

Patsnap is a global innovation intelligence platform covering patents, scientific literature, chemistry, life sciences, and other R&D data.

To explore Patsnap data interactively, try Eureka, Patsnap's AI-native R&D assistant. To access data programmatically, use the MCP servers or REST API via Patsnap Open Platform.

What It Does

Patent & Literature Search MCP gives your AI agent direct access to Patsnap's patent and academic-literature databases. A single natural-language research request can cover both corpora and combine the results for AI reasoning.

  • Patent search — keyword, assignee, inventor, IPC class, legal status, date range, and citation filters
  • Literature search — semantic and keyword search across peer-reviewed scientific papers
  • Cross-domain research — orchestrate patent and paper searches using the same tool and combine the structured results for AI reasoning
  • Natural language input — describe your problem or topic in plain English or Chinese; the MCP constructs the search strategy
  • Precision filtering — filter by assignee, inventor, legal status, jurisdiction, date range, and citation count simultaneously

Data Coverage

This MCP is backed by Patsnap's global patent and scientific-literature coverage.

Patent coverage

Data-source groupRepresentative sourcesCoverage
International and regional authoritiesWIPO/PCT, EPO, ARIPO, EAPO, OAPI, GCCApplications, grants, and regional records
Major national authoritiesUSPTO, CNIPA, JPO, KIPO, DPMA, INPI, UKIPO, CIPO, IP AustraliaApplications, grants, designs, and utility models where applicable
Additional jurisdictionsPatent authorities across Europe, Asia-Pacific, the Americas, the Middle East, and AfricaBibliographic records and authority-supported document types

Patent data coverage includes:

Data typeWhat it covers
BibliographyPublication and application identifiers, titles, dates, applicants, assignees, inventors, and classifications
Full textAbstracts, claims, descriptions, and other document text
PDFOriginal patent-document coverage for supported authorities
Legal eventsStatus changes and patent lifecycle events
File wrapperAvailable prosecution and examination records
LicenseAvailable patent licensing records
LitigationAvailable patent litigation records

Coverage is refreshed frequently, with patent data updated daily.

Scientific-literature coverage

Source typeRepresentative data sources and publications
Biomedical and clinical literaturePubMed, The Lancet, BMJ, JAMA, Blood, Cancer Research, Journal of Clinical Oncology, Gastroenterology
Multidisciplinary journalsNature, Science, Scientific Reports, PLOS ONE, Proceedings of the National Academy of Sciences
Chemistry and materials scienceJournal of the American Chemical Society, Angewandte Chemie, Journal of Chemical Physics, Chemical & Engineering News Archive
Physics and engineeringPhysical Review Letters, Physical Review B, Journal of Applied Physics, AIP Conference Proceedings, SPIE Proceedings
Preprints and repositoriesarXiv, HAL, SSRN, Zenodo, DOAJ
Research datasets and specialist sourcesENCODE Datasets, Worldwide Protein Data Bank, OSTI, RePEc, GBIF, PsycEXTRA

Literature records include bibliographic metadata, abstracts, citation metadata, and PDF coverage. The collection spans journal articles, preprints, conference proceedings, repositories, datasets, and other non-patent literature.

See Patsnap Global Data Coverage for the current authority-level and publication-level coverage matrix.

What You Can Search

patsnap_search searches either patents or scientific literature in one call. Its current input supports:

InputWhat it does
semantic_querySearches a natural-language technical description or concept
search_strategyCombines semantic, keyword, and filter strategies
keywordsRuns precise keyword and BM25 matching
sourceSelects the patent or paper corpus
filtersApplies provider-supported structured filters
sortSorts by relevance or a provider-supported field
offset, limitPaginates results; up to 100 results per call

Patent filters may include assignees, inventors, IPC classifications, jurisdictions, filing or publication dates, legal status, and citation conditions. Exact filter names and operators depend on the current hosted tool schema.

What You Can Retrieve

patsnap_fetch retrieves up to 100 selected result URLs or patent publication numbers as Markdown.

Patent modules

ModuleContent that may be returned
basicBibliographic data, title, abstract, claims, description, and other available record content
citationPatent citations and non-patent literature references
legalAvailable patent legal data and events
familyAvailable patent-family information
include_imagesSigned patent-drawing URLs when available

Citation data

The citation module can retrieve citation data for selected patents. Where available, this may include:

  • patents cited by the selected patent;
  • patents citing the selected patent;
  • applicant, examiner, or search-authority citations;
  • citation categories such as X, Y, and A;
  • cited non-patent literature; and
  • DOI and other bibliographic metadata for cited literature.

Citation direction and type should be determined from the returned record labels. Availability and completeness vary by jurisdiction and patent.

Literature records

Literature records support the basic module and may include title, abstract metadata, authors, affiliations, publication, publication date, DOI, and other available bibliographic identifiers.

Quick Start

1. Get your API key

Register at Patsnap Open Platform and generate a free API key from the Patent & Literature Search MCP page. New accounts include 10,000 free credits. No credit card required.

2. Add to Claude Code

claude mcp add --transport http search-tool \
  "https://connect.patsnap.com/2b0355/logic-mcp?apikey=YOUR_API_KEY"

Restart Claude Code and type /mcp to confirm it loaded.

3. Add to Cursor or Windsurf

Add the following to your MCP configuration file:

{
  "mcpServers": {
    "patsnap_patent_literature": {
      "url": "https://connect.patsnap.com/2b0355/logic-mcp?apikey=YOUR_API_KEY",
      "type": "streamableHttp"
    }
  }
}

Replace YOUR_API_KEY with your key from Patsnap Open Platform.

Run the Containerized MCP Bridge

The repository includes a local stdio bridge for environments that install or deploy MCP servers from source. It exposes static tool definitions without credentials and forwards tool calls to the hosted Patsnap Streamable HTTP service when PATSNAP_API_KEY is configured.

Docker

docker build -t patsnap-patent-literature-search .
docker run --rm -i \
  -e PATSNAP_API_KEY=YOUR_API_KEY \
  patsnap-patent-literature-search

MCP client configuration:

{
  "mcpServers": {
    "patsnap_patent_literature": {
      "command": "docker",
      "args": [
        "run",
        "--rm",
        "-i",
        "-e",
        "PATSNAP_API_KEY",
        "ghcr.io/patsnap/patent-literature-search-mcp:latest"
      ],
      "env": {
        "PATSNAP_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

The GitHub Actions workflow publishes version tags such as v1.1.0 to GHCR as both 1.1.0 and latest.

An optional live upstream check can be run manually from GitHub Actions. Add PATSNAP_TEST_API_KEY as a repository Actions secret, start the workflow with Run workflow, and enable run_upstream. Pull requests never receive or use this secret.

Tools

  • patsnap_search — Search either patents or scientific literature in a request. Supports natural-language and semantic queries, precise keyword and BM25-based text search, and filters such as assignee, inventor, IPC classification, legal status, jurisdiction, date range, and citations. A single research request may invoke patent and literature searches and combine their results.
  • patsnap_fetch — Retrieve a patent or literature record as Markdown using one or more result URLs. Patent records can also be fetched by publication number and may include bibliographic data, claims, descriptions, drawings, citations, legal data, and patent-family data. Literature records include core bibliographic and abstract metadata.

Development

Requires Node.js 20 or newer.

npm ci
npm test
npm start

The test suite verifies:

  • static tool discovery without an API key;
  • argument validation and upstream forwarding;
  • structured MCP error handling;
  • a real stdio initialize and tools/list handshake.

To verify the container locally:

docker build -t patent-literature-search-mcp:test .
npm run test:docker

The server writes MCP JSON-RPC only to stdout. Application errors are returned as MCP tool results, and credentials are never included in logs.

Glama Release

After claiming the server on Glama:

  1. Open the server's Dockerfile Admin page.
  2. Select the repository-root Dockerfile.
  3. Add PATSNAP_API_KEY as a secret deployment environment variable.
  4. Use node src/index.js as the command if Glama does not infer the Dockerfile CMD.
  5. Run Deploy, verify the MCP build test, and create a release.

Example Prompts

What is Tesla actually working on in battery technology — show me
their recent patent filings, group by technology area, and identify
who is leading each.
Which GLP-1 receptor agonist patents are expiring before 2028,
and what does recent literature say about next-generation mechanisms?
My EV battery loses 40% range in cold weather. What technical
solutions exist in patents and academic papers?

Resources

License

Apache-2.0


Powered by Patsnap. Innovate with Confidence.

Files in the repo

Repository payload15 top-level entries
  • .github
  • scripts
  • src
  • test
  • .dockerignore
  • .env.example
  • .gitignore
  • Dockerfile
  • glama.json
  • LICENSE
  • package-lock.json
  • package.json
  • README.md
  • server.json
  • SKILL.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More connectors

Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

86k

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

43k

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code

14k
okf-memory/
okf-agent-memory

Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.

547
tirth8205/
code-review-graph

Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.

31k
2akouwu/
reverify

Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.

1.1k