Why this needs Advanced Patent Search

The core difference is semantic, not just structural: this tool matches on the technical meaning behind a description rather than requiring the exact keywords or Boolean syntax a traditional patent search demands. Advanced Patent Search adds 17 tools to your AI assistant or coding agent — semantic, similar-patent, image, patent-number, nested, and keyword-assist search among them — drawing on the same corpus behind Patsnap’s other tools: 210M+ patent records across 174 jurisdictions, updated on a rolling basis. Patsnap has built that kind of search specifically for patents since 2007, which is a meaningfully different starting point than pointing a general-purpose search model at patent text it wasn’t tuned for.

Prepare your query

The semantic search tool accepts a text description of the technical area you want to explore — the tool recommends a passage longer than 200 words for best results, though shorter descriptions still return ranked matches. For this tutorial, the input is a short tutorial-length example: battery recycling lithium recovery methods. In a real workflow, you’d supply a fuller description from your invention disclosure, research summary, or technical specification, covering the problem, the proposed solution, and any distinguishing materials or process steps.

1Connect Advanced Patent Search

1. Get a Patsnap Open Platform API key.
2. Copy the connection URL or JSON config from the Connect panel on the Advanced Patent Search page.
3. Add the server to your AI assistant or coding agent or another MCP-compatible client, then restart it.
4. Confirm the full set of 17 search tools is available before you ask.

2Run the example query

Search for patents matching "battery recycling lithium recovery methods" and return the top result.

Claude calls search_patents_by_semantic(text="battery recycling lithium recovery methods", limit=1). The limit parameter controls how many results come back — set to 1 here so you can review the structure of a single match before requesting more. The tool runs a semantic search across Patsnap’s patent collection and returns results ranked by relevancy, a similarity signal for how closely each document matches the input description.

3What the result tells you

This single record gives you the identifiers you need to open the full patent, review its claims and description, and decide whether it’s relevant to your research question. Because the example used limit=1, only one result came back — in a real search, set a higher limit, such as 10 or 20, to review a broader set of matches and compare technical approaches. A fuller, 200+ word description than the tutorial example will also help the semantic engine rank results more precisely.

4Go further with follow-up questions

From the same session, Ask your AI assistant or coding agent to:

• Retrieve more matches — repeat the search with a higher limit, such as limit=10, to compare top-ranked patents and spot common assignees or alternative technical approaches.
• Request specific fields — use the returned publication number to ask for claims, description, images, or legal status. “Show me the claims for WO2024165621A1” calls the right tool directly.
• Adjust the description — if results are too broad or narrow, add distinguishing process steps, materials, or performance targets and re-run the search to see how the ranking shifts.

Each follow-up builds on the structured evidence already returned, so you can move from discovery to detailed review without switching platforms or re-entering publication numbers.