The Patsnap API is a set of programmatic interfaces available through Patsnap Open for accessing patent, research literature, life sciences, and related innovation data. Developers use these interfaces to put selected records into an application, research pipeline, or AI workflow instead of repeatedly copying information from a browser.
The practical starting point is the information your application needs to return. A search box needs discovery results; a company research screen needs identifiable records; an AI assistant needs evidence it can cite. Patsnap Open provides data APIs and separate AI capabilities, but each operation has its own inputs, response fields, and access requirements. The API reference is the place to check those boundaries.
What can you access through Patsnap Open?
Start with the entity you need, then select an operation. The catalog includes the following areas; this is a map of the catalog, not a promise that one endpoint or subscription returns every field.
| Area | Application question | What to verify before building |
|---|---|---|
| Patent and IP data | Which patent records match a topic or known identifier? | Search fields, bibliographic fields, family grouping, and separate legal or citation operations |
| Science and technology | Which research records help investigate a technical question? | Literature coverage, available metadata, abstract availability, and links to source material |
| Life sciences | Which drug, trial, or biological records fit the research task? | Entity identifiers, endpoint scope, permitted uses, and the meaning of each returned field |
| AI capabilities | Can a documented AI operation assist with a defined analysis task? | Required evidence, output format, task limits, and what a reviewer must still check |
A successful response means the operation ran. It does not establish that your research question has been fully answered. For example, a bibliographic record can identify a patent without supplying every claim, ownership event, or jurisdiction-specific status needed for a later review.
Which access method should you use?
Choose the access method around how the work will run. A developer testing fields has different needs from an analyst exploring a topic or a team maintaining a scheduled integration.
| Access method | A useful starting point when… | Your responsibility |
|---|---|---|
| API Explorer | You want to inspect a documented request and response interactively | Check the chosen endpoint and avoid pasting secrets into shared examples |
| REST API | Your backend needs repeatable retrieval and explicit application logic | Handle authentication, errors, pagination, storage, and usage controls |
| MCP | A compatible AI application should call available tools during a conversation | Check client support, permissions, exposed tools, and source handling |
| Browser research interface | A person needs to explore and read records directly | Preserve the search scope and source references when handing work to others |
REST gives your application control over the HTTP calls. MCP makes tools discoverable to compatible AI clients. These can serve the same research task at different stages; they are not interchangeable response formats. The Patsnap MCP catalog describes the available servers and their intended tasks.
How authentication works
Patsnap documents API key authentication using an Authorization header with the Bearer scheme. Requests go over HTTPS; the REST base URL is https://connect.patsnap.com. JSON is used for request and response data. Follow the authentication guide and the REST overview for the current conventions.
Create a key in your account, save it securely, and load it from a server-side secret store or environment configuration. Keep it out of public source code, browser JavaScript, screenshots, prompts, and logs. Use a separate development key so testing and production access can be managed independently.
Authentication and data access are different checks. If a request fails, inspect the documented error and account permissions before changing the query. A missing field may reflect the endpoint or record; it is not automatically an authentication problem.
Make your first request
Use one small question whose answer you can inspect manually. For example, test whether a patent search returns records relevant to a familiar technical phrase. The first goal is to understand the response, not to collect a large dataset.
- Create and securely store an API key using the documented account process.
- Select a specific endpoint in the API reference. Read its required inputs and response schema before using an example from another endpoint.
- Run a small request in API Explorer. For patent query search, inspect the documented
query_textinput and grouping options. - Read the response as data: identify the record IDs, relevant fields, any paging information, and missing values.
- Open a representative source record and check that your application interprets its identifier and dates correctly.
- Move the tested request into your backend, then add controlled error handling and usage logging without recording secrets.
The patent query-search reference is a useful example of an operation-specific contract. Its settings should not be copied blindly to a literature or life sciences endpoint. Use the first-request guide for the connection steps.
How to evaluate whether the API fits
Before choosing an integration, test a representative set: a known record, an ambiguous query, a record with missing fields, and a search likely to return several pages. Write down what the application must show when the data is incomplete.
| Evaluation question | Evidence to collect |
|---|---|
| Does the source cover the task? | Relevant jurisdictions, document types, dates, languages, or scientific domains |
| Can records be joined reliably? | Identifier definitions, normalization rules, and cross-source matching checks |
| Can the workload run predictably? | Endpoint limits, pagination behavior, account permissions, and current usage terms |
| Can a user inspect the result? | Source references, retrieval time, field meanings, and retained search settings |
| Can the data be used as intended? | Applicable storage, redistribution, licensing, and access conditions |
Evaluate public and specialist sources against the same requirements. A narrow official dataset may be sufficient for a jurisdiction-specific task. A broader commercial service may reduce integration work when several record types are needed. Neither choice removes the need to verify field meanings and important conclusions.
Using Patsnap APIs with AI applications
Keep retrieval and interpretation visible as separate steps. A backend can call REST endpoints, select relevant fields, and pass those records to a model with instructions to cite identifiers and flag uncertainty. With MCP, a compatible client can discover and invoke available tools as part of the conversation.
For Claude, Codex, or ChatGPT, check the specific application, version, plan, and administrator settings before assuming that an MCP connection is available. A model name alone does not establish connector support. Begin with a read-only research question, inspect the tool result, and confirm that the final answer cites retrieved evidence rather than inventing missing details.
Make your first Patsnap API request
Use the official quickstart to connect your account and inspect a small response before integrating it.
Frequently asked questions
What is the Patsnap API?
It is the programmatic access offered through Patsnap Open for innovation data and documented capabilities. Choose individual endpoints according to the records and operations your application needs.
How do I authenticate?
Use an API key according to the current authentication guide, preferably in the Bearer authorization header. Keep the key on the server side and use HTTPS.
Should I choose REST or MCP?
Choose REST when your code should control the request sequence. Consider MCP when a compatible AI client should discover and call tools. In both cases, review the returned sources.
Can it support an AI application?
Yes, retrieved records can provide evidence for an AI application. Your system still needs instructions, source attribution, access controls, and checks for unsupported interpretations.
Where are endpoints documented?
The Patsnap Open API reference contains operation-level documentation. Confirm the selected endpoint's method, inputs, fields, and current access requirements before implementation.
Sources and disclosure
Product descriptions are based on the official Patsnap documentation linked above, checked on September 17, 2026. This article is published by Patsnap. The request sequence and evaluation checklist are implementation guidance, not results from a production benchmark. Endpoint availability, permissions, and terms should be checked for your account.