Patsnap Ip Searching

Setup

Get your API key at https://open.patsnap.com

Patsnap Patent Research

This skill connects your AI agent to Patsnap's Patent Research MCP server.

Prerequisites

{
  "mcpServers": {
    "patsnap_patent_research": {
      "url": "https://connect.patsnap.com/f176d7/mcp?apikey=YOUR_API_KEY",
      "type": "streamableHttp"
    }
  }
}

For the full tool schema, refer to: https://open.patsnap.com/marketplace/mcp-servers/patsnap-ip-searching

Instructions for AI Agents

Choose the correct workflow

• Use novelty_search for a technical solution, invention disclosure, or claim-text novelty review. Text is the primary input; images are optional supplements. Optional search channels include semantic, Boolean, paper, and web search.
• Use fto_review for invention-patent infringement-risk research based on a product or technical implementation. The required input object uses text as its primary content and can include translated text and images.
• Use design_fto for design-patent risk research. Its required input object must contain at least one image; only the first item in input.images is used, and a title is optional.
• Use a general patent-search MCP instead when the user needs open-ended discovery, query construction, or retrieval without a novelty or FTO analysis workflow.

Configure search and filters carefully

For fto_review, search.mode supports lite for rapid screening and pro for full analysis. Its search settings can also control candidate-patent count and iterative search rounds. Optional filters support country, apd, legal_status, and assignee; application-date ranges use YYYYMMDD, and assignee filters support not_in.

For design_fto, optional filters support country, apd, legal_status, and loc; loc uses Locarno classifications and application-date ranges use YYYYMMDD.

The analysis tools also accept optional execution settings for asynchronous execution or real-time progress, optional output controls such as normalized result-row limits, and an optional caller-supplied task_id.

Retrieve asynchronous results

When a tool runs asynchronously, preserve its returned task_id and call get_task. The task_id is required. Use view: overview for the status summary; this is the default. Use view: detail when step and result details are needed. Do not start duplicate analyses merely because a task is still running.

Interpret results responsibly

Preserve the user's technical scope, countries, date window, legal-status assumptions, and other filters in the report. Distinguish retrieved patents and service-generated comparisons from your own interpretation. Novelty and FTO outputs support research and triage; they do not establish patent validity, non-infringement, or a legal opinion. Recommend qualified patent counsel for decisions carrying legal or commercial risk.

Example Workflows

• Search the novelty of an invention disclosure and compare its extracted features with candidate patents.
• Run a lite FTO screening, then use pro analysis for the highest-risk implementation.
• Restrict an invention FTO review by country, application date, legal status, and excluded assignees.
• Review a single product image for design-patent risk in specified Locarno classes.

Resources

• MCP Server: Patsnap Patent Research
• Patsnap Open Platform: open.patsnap.com
• Patsnap: https://www.patsnap.com