Patent Research MCP
An MCP server for patent research and enterprise architecture pattern extraction.
README
patent-research-mcp
A generic MCP server for patent research, architecture extraction, claims analysis, and pattern synthesis.
Fetches patents from Google Patents, extracts structured sections, analyzes architectural patterns, assesses claims risk, and synthesizes reusable design patterns — all through Model Context Protocol (MCP) tools.
Why
Patent research for system architecture is repetitive and error-prone. This tool automates the pipeline:
fetch → extract → analyze → firewall → pattern → export
Each step produces structured, reviewable artifacts that feed into architecture decision records (ADRs), not unstructured notes.
Quick Start
# Install with dev dependencies
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
# Verify
patent-research seeds
CLI Usage
| Command | Description |
|---|---|
patent-research seeds |
List seed patents with metadata |
patent-research fetch <number> [--pdf] |
Download patent HTML + plain text |
patent-research sections <number> |
Extract structured sections (abstract, claims, description, etc.) |
patent-research patterns |
List saved architectural patterns |
patent-research export |
Generate complete research summary |
# Complete workflow
patent-research seeds
patent-research fetch US7979296B2
patent-research sections US7979296B2 --save
patent-research patterns
patent-research export
MCP Server
Run as a standard MCP server for use with any MCP client (Claude Desktop, Hermes, etc.):
python -m patent_research_mcp.server
Tools (10 total)
| Tool | Description |
|---|---|
patent_seed_list |
List all seed patents with metadata |
patent_fetch |
Download patent HTML, plain text, and optional PDF |
patent_get_sections |
Extract structured sections from a fetched patent |
architecture_card_save |
Save a structured architecture analysis |
claims_firewall_save |
Save a claims risk assessment |
pattern_save |
Save a reusable architecture pattern |
pattern_list |
List all saved patterns |
pattern_compare |
Compare patterns across patents for shared concepts |
research_export_markdown |
Generate a complete research summary |
module_proposal |
Generate a module proposal template |
Example: Hermes config
mcp:
servers:
patent-research:
command: /path/to/.venv/bin/python
args: [-m, patent_research_mcp.server]
enabled: true
env:
RESEARCH_PLUGIN: /path/to/private-plugin # optional
Data Flow
Google Patents
│
▼
fetch() ─── raw/ (HTML + TXT + optional PDF)
│
▼
get_sections() ─── sections/ (structured JSON)
│
▼
ArchitectureCard ─── cards/ (architecture analysis)
ClaimsFirewall ─── claims/ (risk assessment)
PatternCard ─── patterns/ (reusable patterns)
│
▼
export() ─── exports/ (research summary)
Artifacts
| Artifact | Schema | Purpose |
|---|---|---|
| ArchitectureCard | ArchitectureCard |
Structured architecture analysis: problem, components, ontology, patterns |
| ClaimsFirewall | ClaimsFirewall |
Risk assessment: dangerous claims, safe abstractions, design-around |
| PatternCard | PatternCard |
Reusable pattern: entities, events, states, reusable principle |
Plugin System
The server supports private plugins via the RESEARCH_PLUGIN environment variable. When set, the plugin's patents.json overrides the default seed list. This allows domain-specific patent collections without forking the generic core.
patent-research-mcp/ # public, generic
└── src/patent_research_mcp/
└── server.py # checks $RESEARCH_PLUGIN/patents.json
private-plugin/ # private, domain-specific
└── patents.json # SeedPatent array
Example patents.json:
[
{
"publication_number": "US20220237532A1",
"title": "Digital Twin of Organizational Processes",
"domain": "enterprise-architecture",
"why_it_matters": "...",
"google_patents_url": "https://patents.google.com/patent/US20220237532A1/"
}
]
Schemas
| Class | File | Description |
|---|---|---|
ArchitectureCard |
schemas.py |
Full patent architecture analysis |
ClaimsFirewall |
schemas.py |
Claims liability assessment |
PatternCard |
schemas.py |
Reusable architectural pattern |
SeedPatent |
schemas.py |
Seed patent entry |
CompareResult |
schemas.py |
Cross-pattern comparison result |
Environment Variables
| Variable | Default | Description |
|---|---|---|
PATENT_RESEARCH_DATA |
$CWD/data |
Data storage directory |
RESEARCH_PLUGIN |
— | Path to private plugin with custom seeds |
Development
# Install dev dependencies
pip install -e ".[dev]"
# Run tests
pytest -v
# Lint
ruff check src/
ruff format --check src/
# Type check (optional)
mypy src/
Project Structure
patent-research-mcp/
├── src/patent_research_mcp/
│ ├── __init__.py
│ ├── server.py — MCP server (tools) + CLI
│ ├── schemas.py — Pydantic models
│ ├── patents.py — Google Patents fetcher (Playwright)
│ ├── store.py — JSON file storage
│ ├── normalizer.py — Text & synonym normalization
│ ├── exporter.py — Markdown report generation
│ └── seed.py — Default seed patent data
├── prompts/ — LLM prompt templates for analysis
├── data/ — Patent artifacts (gitignored except examples)
├── tests/ — pytest suite
└── pyproject.toml — Build config
License
MIT — see LICENSE for details.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。