Public IP data for agents¶
Patent Client Agents is an open-source Python library and MCP server for researching patents, trademarks, designs, copyright records, IP law, published decisions, and official fee schedules. It exposes source-specific clients as typed tools without hiding where a record came from.
Try the hosted service Install an agent plugin Browse the source atlas
Choose the right way to run it¶
| Use | Best for | What you control |
|---|---|---|
| Hosted MCP | A first query with no local setup | Google sign-in and a deliberately curated public surface |
| Agent plugin | Private research in Claude Code, OpenAI Codex CLI, or Google Antigravity CLI | The same Satchel-generated package in each client; 136 tools by default, up to 234 when every gated family is configured |
| Stdio MCP server | Other MCP clients or a hand-managed local configuration | The server command, environment, and client configuration |
| Python library | Application code and source-specific integrations | The exact clients, credentials, caching, and runtime your application needs |
The hosted service is convenient, but it is not the full product surface and should not receive confidential material. The agent plugins are the shortest path to a private local setup in Claude Code, Codex, or Antigravity. Configure the stdio server directly for other MCP clients or when you need to control the command and environment yourself.
Start from a research task¶
Find and compare patent records¶
Use global search to identify relevant publications, then retrieve claims, citations, families, legal events, or prosecution records from the appropriate office source.
- Google Patents for broad discovery and document data
- USPTO Publications for U.S. full-text search
- USPTO Applications for application records and file history
- EPO OPS for European and worldwide bibliographic, family, and legal-event data
Reconstruct prosecution and ownership¶
Follow application events, office actions, cited references, recorded transfers, and agency proceedings without treating any single database as a complete file.
Research law, guidance, and proceedings¶
Search examiner manuals, statutes, published opinions, and public tribunal data through clients that retain each source's native citations and identifiers.
- MPEP, TMEP, and WIPO Lex
- Federal Circuit opinions
- USITC EDIS, DataWeb, HTS, and IDS
- Japan IP High Court and China SPC IP Court
Understand coverage before you depend on it¶
Coverage is intentionally explicit. The Patent Client Index records each data product, its status, rights, data types, access conditions, and source documentation. The public atlas adds office-level research:
- Green means a usable programmatic path exists.
- Yellow identifies a credential or paid-access condition.
- Red records a legal or technical reason a connector is not shipped.
These ratings describe access, not the completeness or legal sufficiency of an individual research result. Verify material conclusions against the linked official record.
What a client provides¶
Each integration follows the upstream source rather than forcing every system into one lowest-common-denominator schema. Depending on the source, a client may provide typed models, source URLs, retrieval details, caching, rate limiting, retry behavior, normalized errors, or local searchable corpora.
This design preserves distinctions that matter in IP research: publication and application numbers, family relationships, office-specific status fields, effective dates, court identifiers, and the difference between a register, search index, bulk dataset, and published document.
Next steps¶
- Install the shared agent plugin in Claude Code, Codex, or Antigravity.
- For another client, configure the local MCP server.
- Find the source you need in the Patent Client Index.
- Open its page under API Reference for exact methods, models, limits, and examples.
The project is developed in public on GitHub under the Apache-2.0 license.