How to Search Patents and Scientific Literature in One Claude Workflow
R&D researchers working on competitive technology analysis or drug development typically need to search two separate domains: patent databases for proprietary disclosures and scientific literature for published research. When you ask Claude to find connections between GLP-1 receptor agonist patents and related academic papers, Claude cannot retrieve current technical documents on its own. It relies on what was known at training time and cannot access live patent or literature databases.
Patsnap Patent & Literature Search bridges this gap by connecting Claude to both patent and scientific literature sources through a single MCP server. You ask Claude one natural-language question and the MCP server retrieves current results from both domains, returning them as structured Markdown you can review directly in the conversation.
This tutorial shows you how to prepare a combined query, connect Patsnap Patent & Literature Search to Claude Desktop, run a dual-source search, and interpret the returned evidence.
Why use Patsnap Patent & Literature Search for this search in Claude
Generic LLMs produce suggestions and summaries but cannot execute live queries against proprietary patent and literature databases. When you need the actual publication number of a recent filing or the DOI of a specific meta-analysis, you require access to curated, indexed records that exist outside the training corpus.
Patsnap Patent & Literature Search searches patents and scientific literature from one MCP entry point, translating natural-language queries into structured database requests and returning Markdown-formatted results directly in Claude. You write your research question in plain English, and the MCP server retrieves current hits from both domains, complete with titles, identifiers, and relevance ranking.
Prepare the research question and required input
Before you connect, formulate your research question and identify the key technical terms or concepts. For a combined patent and literature search, choose terminology that appears in both patent claims and academic abstracts. Avoid overly narrow jargon that only one domain uses, and prefer functional or chemical descriptors that translate across contexts.
For this tutorial, the standard case uses ["GLP-1", "receptor agonist"] with a dual-source request. "GLP-1" is a well-known therapeutic target, and "receptor agonist" describes a functional mechanism documented in both proprietary filings and published pharmacology papers. This pair of terms will retrieve patents describing GLP-1 agonist synthesis or formulation and scientific papers discussing GLP-1 receptor pharmacology or clinical trials.
Write your terms as a short list of strings. You will pass them to the MCP server along with a source parameter that directs the search to patents, papers, or both.
1Connect Patsnap Patent & Literature Search to Claude
The official MCP flow is: obtain a Patsnap Open Platform API key, choose a server, generate a connection link, and configure a compatible client such as Claude Desktop. Patsnap Open Platform’s Starter plan is free and includes 10,000 credits for initial testing.
Once connected, Claude can call the patsnap_search tool with your keyword list and source parameter. The MCP server handles authentication, query translation, and result formatting, returning structured Markdown records to the Claude conversation.
connect.patsnap.com/2b0355/logic-mcp?apikey=yourapikey- Copy the connector address above.
- Open your MCP client (for example Claude Desktop or Claude Code) settings and add a new MCP server.
- Paste the address and replace
yourapikeywith your Patsnap Open API key. - Save and reconnect; the new tools become available in your next conversation.
2Run the example query
Ask Claude to search for patents and papers on GLP-1 receptor agonists. Claude will invoke the MCP server with the dual operation: first searching patents, then searching scientific literature, both using the same keyword pair so you can compare coverage and top-ranked results across the two domains.
The query retrieves current results from both patent and scientific databases using natural-language support and returns them in Markdown format for direct review.
3What the search result gives you
The MCP server returns two result sets. The patent search found 943,453 hits, and the top result was US20220153804A1, a filing by Enzene Biosciences describing a GLP-1 synthesis process. The scientific literature search found 202,014 hits; the top result was a paper on glp-1 and Notch gene interactions, which is unrelated to therapeutic receptor agonists.
Each result includes the document identifier (patent publication number or DOI), title, assignee or author, and a short context snippet. The hit counts show the size of the retrieved set, and the ranking reflects the database’s relevance scoring. You can scan the top results to identify high-priority documents for full-text review.
The difference in top-ranked results illustrates why combined searches matter: the same keywords retrieve distinct document types, and cross-domain review helps you spot whether a patent-holder’s technical approach has corresponding published validation or whether academic findings have not yet been claimed in a filing.
4Refine the search with focused follow-up questions
After reviewing the initial results, refine your search with targeted follow-up questions that use identifiers or metadata already returned. For example:
- Ask Claude to search patents assigned to Enzene Biosciences to find related filings by the same organization.
- Narrow the keyword list by adding a therapeutic indication such as "diabetes" or "obesity" alongside "GLP-1 receptor agonist."
- Pivot to a thematic search by asking for patents and papers related to a specific mechanism or formulation detail mentioned in the returned snippets.
Each refinement reuses the same MCP connection, and Claude translates your natural-language request into the appropriate source and keyword parameters. These follow-up searches help you map the boundary between proprietary disclosures and published research, identify white space, and prioritize documents for detailed technical review.
Frequently asked questions
Can Claude search patents and scientific literature without an MCP connection?
No. Claude’s training data does not include live access to patent or scientific literature databases. To retrieve current patent publication numbers, DOIs, and structured records, Claude requires an MCP server such as Patsnap Patent & Literature Search that connects to curated, indexed sources.
What input format does Patsnap Patent & Literature Search accept?
The MCP server accepts a list of keyword strings and a source parameter. For a combined search, you provide the same keyword list with source="patent" for the patent database and source="paper" for the scientific literature database.
How many results does the search return?
The MCP server returns the total hit count and ranks the results by relevance. In the standard case, the patent search found 943,453 hits and the literature search found 202,014 hits. You review the top-ranked results and can refine the query to narrow the set.
What does the returned result include?
Each result includes the document identifier (patent publication number or DOI), title, assignee or author, and a short context snippet. You use these fields to identify high-priority documents for full-text review.
Can I refine the search after seeing the initial results?
Yes. You can ask Claude to search patents by a specific assignee, narrow the keyword list by adding a therapeutic indication, or pivot to a thematic search based on mechanisms or formulation details mentioned in the returned snippets. Each refinement reuses the same MCP connection.
Use Patsnap Patent & Literature Search in Claude
Connect the MCP server and continue this workflow with approved Patsnap data.
