By the Patsnap Open Team · Updated August 2026

This workflow works in Claude, Claude Code, ChatGPT, or Codex when the selected client supports the required MCP connection. After connecting, the research steps and evidence boundaries are the same.

When you have a known compound structure and need to identify where it appears in the patent literature, you face a retrieval task that general language models cannot complete alone. An AI assistant or coding agent can help you articulate the search objective and interpret molecular data, but it has no access to chemical structure databases or patent collections on its own.

Chemical Molecular is an MCP server that supplies structure-based compound search across patent records. It accepts standard molecular identifiers such as SMILES strings and returns matching patent-disclosed structures along with physicochemical identifiers and first-publication references. By connecting this server to your AI assistant or coding agent, you can ask natural-language research questions and retrieve current structure-to-patent evidence in one workflow.

This tutorial walks through a structure search using aspirin as the example compound, shows what the returned match data looks like, and demonstrates how to refine your search with follow-up questions.

1Why use Chemical Molecular for this search with AI

Generic language models do not index chemical structures or patent documents, so they cannot perform a molecular search or tell you which patents disclose a given compound. Chemical Molecular bridges this gap by offering structure-led retrieval across patent chemistry records. You supply a SMILES string or similar identifier, and the server returns matched structures with their InChIKey, molecular formula, and first-publication patent numbers.

This structure-based search is especially valuable when you start from a known molecule rather than a text keyword, allowing you to find exact or extended molecular matches even when compound names vary across documents or jurisdictions.

2Prepare the research question and required input

Before you query, identify the compound you want to search and obtain its SMILES representation. SMILES (Simplified Molecular Input Line Entry System) is a compact text notation that encodes molecular structure. You can generate a SMILES string from drawing tools, chemical databases, or published literature.

For this example, the aspirin SMILES is CC(=O)OC1=CC=CC=C1C(=O)O. You will also choose a match type: exact (EXA) matches the structure precisely, while extended (EXT) may include derivatives or closely related molecules that require expert review to assess relevance.

Prepare your question in natural language, such as "Find patents containing this chemical structure," and have the SMILES string ready to supply when Claude asks for it.

3Connect Chemical Molecular to your AI workflow

1. Get a Patsnap Open Platform API key from your account.
2. Open the Chemical Molecular Connect page above and copy the connection URL or configuration shown there.
3. Add the server to your AI assistant or coding agent or another MCP-compatible client, then restart or reconnect the client.

Once connected, Ask your AI assistant or coding agent to search for the supplied structure in plain language. You do not need to enter the internal tool name unless you are debugging the integration.

4Run the example query

Ask your AI assistant or coding agent to search for patents containing the aspirin structure. Provide the SMILES string CC(=O)OC1=CC=CC=C1C(=O)O and specify an extended match type with a result limit of one for this demonstration.

Claude calls the ls_structure_search tool with these parameters:

smiles: CC(=O)OC1=CC=CC=C1C(=O)O
type: EXT
limit: 1

The tool retrieves matching structures from the patent chemistry database and returns the requested number of representative results.

5What the observed search returned

The recorded extended-match query reported 1,396 matched structures. One representative record from that result set is shown below.

6Refine the search with focused follow-up questions

Once you have initial results, you can refine the search by adjusting the match type, increasing the result limit, or exploring specific returned identifiers. For example:

• "Show me the next ten matched structures from the same search" uses the same SMILES and parameters but raises the limit to retrieve additional candidates.
• "Fetch detailed properties for InChIKey XZSTVWKWQAULJQ-UHFFFAOYSA-N" retrieves the full molecular record for a specific match.
• "Perform an exact match search with the same SMILES" switches to type=EXA to narrow results to structures identical to the query.

Each follow-up builds on the identifiers and parameters already returned, letting you explore the match set systematically or pivot to narrower criteria without starting over.