Patents MCP Server | Documentation

Overview

The server exposes 4 tools in two groups. Use search to find patents, then use the lookup tools to fetch the full detail for the ones you care about.

Add the server URL above to any MCP-compatible client as a remote server. Your assistant then chooses the right tool from the descriptions below. Some clients show tool names with an ipone-patents___ prefix.

Workflows

Find and review patents

1. Search

Run a fielded query. Request only the fields you need, and a listref if you may page through results later.

searchByQuery
2. Retrieve

Fetch fuller records for chosen patents: abstract, claims, description, family and citations.

documentsByPndocumentsById
3. Page through

Return more pages, or different fields, for an earlier search without running it again.

documentsByListref

Example: fielded search (illustrative values)

{
  "tool": "searchByQuery",
  "arguments": {
    "query": "ti=(lithium ADJ battery) AND cc=US",
    "return_fields": "pn,ti,pa,pd",
    "size": 10,
    "return_listref": true
  }
}

Summarize a result set

Pass summarize to searchByQuery to aggregate results by one or more fields, for example pa,cc for the top assignees and countries. Use summarizeSize (1 to 100) to control how many values come back for each field.

Look up a known patent

If you already have a publication number, skip the search and call documentsByPn with the fields you want.

Example: look up by publication number (illustrative values)

{
  "tool": "documentsByPn",
  "arguments": {
    "pns": "<publication number>",
    "return_fields": "ti,ab,cl,fam,cipf"
  }
}

Tool reference

Return fields differ between the search tool and the lookup tools. See the field reference for which tags each one returns.

Search

searchByQuery

Fielded boolean searchQuery up to 1,000 characters

Searches patents with a fielded query. Supports boolean operators, wildcards, phrase queries and proximity (ADJn, NEARn). Returns the fields you request, and can return a listref for paging or aggregate results with summarize.

Parameter	Type	Required	Notes
query	string	Yes	Fielded query, for example `ti=(lithium ADJ battery) AND cc=US`. Maximum 1,000 characters. See query syntax.
collections	string	No	Comma-separated collection names. Defaults to all collections.
return_fields	string	No	Comma-separated field tags. Default `id`. See the field reference.
size	integer	No	Maximum rows to return. Default 10.
offset	integer	No	Zero-based result offset. Default 0.
return_listref	boolean	No	Return a `listref` that identifies this search.
listref	string	No	A reference to a previously executed query. Leave unset unless you specifically need it.
search_on_datatype	string	No	One of `fld`, `dwpi` or `fld_dwpi`. Default `fld_dwpi`.
summarize	string	No	Comma-separated field tags to aggregate: `ad`, `cc`, `cpc`, `ic`, `in`, `pa`, `pd`, `py`.
summarizeSize	integer	No	Maximum aggregation values per field, 1 to 100. Default 10.

Lookups

documentsByPn

Retrieves patent data for one or more publication numbers.

Parameter	Type	Required	Notes
pns	string	Yes	Patent (publication) numbers.
collections	string	No	Comma-separated collection names. Defaults to all collections.
return_fields	string	No	Comma-separated field tags. Default `pn`.
size	integer	No	Maximum rows to return. Default 10.
offset	integer	No	Zero-based result offset. Default 0.
listref	string	No	A reference to a previously executed query. Leave unset unless you specifically need it.

documentsById

Retrieves patent data for one or more document GUIDs, such as the id values returned by a search.

Parameter	Type	Required	Notes
ids	string	Yes	Document GUIDs.
collections	string	No	Comma-separated collection names. Defaults to all collections.
return_fields	string	No	Comma-separated field tags. Default `pn`.
size	integer	No	Maximum rows to return. Default 10.
offset	integer	No	Zero-based result offset. Default 0.

documentsByListref

Retrieves patent data for a previous search using its listref. Use it to fetch further pages or different fields without running the query again.

Parameter	Type	Required	Notes
listref	string	Yes	The listref returned by an earlier `searchByQuery` call made with `return_listref` set to true.
collections	string	No	Comma-separated collection names. Defaults to all collections.
return_fields	string	No	Comma-separated field tags. Default `pn`.
size	integer	No	Maximum rows to return. Default 10.
offset	integer	No	Zero-based result offset. Default 0.

Field reference

Use the two-to-five letter tag, not the display name, in query, return_fields and summarize. Only tags marked for a use are accepted for it.

Tag	Field	Search	Returned by search	Returned by lookups	Summarize
ti	Title	✓	✓	✓	—
ab	Abstract	✓	—	✓	—
cl	Claims	✓	✓	✓	—
dsc	Description	✓	—	✓	—
pd	Publication date	✓	✓	✓	✓
py	Publication year	✓	—	✓	✓
ad	Application date	✓	✓	✓	✓
ay	Application year	✓	—	✓	—
an	Application number	✓	✓	✓	—
pan	Derwent accession number	✓	✓	✓	—
pn	Publication number	✓	✓	✓	—
cc	Country code	✓	—	✓	✓
id	Document identifier	—	✓	✓	—
pa	Assignee	✓	✓	✓	✓
in	Inventor	✓	✓	✓	✓
copa	Optimized assignee	✓	✓	—	—
copaa	All optimized assignee	—	—	✓	—
cuppa	Ultimate parent	✓	✓	—	—
cuppaa	All ultimate parent	—	—	✓	—
cpc	CPC classification code	✓	✓	✓	✓
ic	IPC classification code	✓	✓	✓	✓
ipc	IPC classification code (alternate tag)	—	✓	—	—
cip	Citing patent	✓	—	✓	—
cipf	Cited patents	—	—	✓	—
cin	Non-patent citations	✓	—	✓	—
fam	Family members	—	—	✓	—
caid	Patent dead/alive status	✓	✓	✓	—
aisq	Semantic (Derwent AI) search, free text	✓	—	—	—

“Lookups” means documentsByPn, documentsById and documentsByListref. The aisq tag runs a free-text semantic search rather than a keyword match.

Query syntax and limits

Fielded queries Write clauses as field=value and combine them with AND, OR and NOT. Phrase, wildcard and proximity searching are supported. When the same field repeats across adjacent AND clauses, prefix it once, as in ti=title AND device AND cc=US.
Multiple terms in one field Wrap the value in parentheses. If you give several terms with no operator between them, put ADJ between them, as in ti=(lithium ADJ battery).
Proximity, wildcard and truncation limits ADJn and NEARn accept n from 1 to 10. Use no more than 10 wildcard characters per term. Truncation is limited to 5 terms per query. The query itself is limited to 1,000 characters.
Collections By default all collections are searched. To narrow the search, pass a comma-separated list of exact collection names. Names are a country code plus a type suffix (apps, grants or utils), for example usapps, epgrants or jputils.
Request only what you need Search returns id by default and the lookup tools return pn. Add fields to return_fields deliberately, and refine broad queries rather than paging through everything.
Rate limits The four tools share one limit of 50 requests per rolling 60 seconds and 50 concurrent requests. If a call is throttled, wait for retry_after_seconds when it is provided instead of retrying immediately.