Privacy policy
What we collect, why, and when it is deleted.
PatentReply is operated by Patent Agentic LLC, a Delaware limited liability company (“we”). This policy covers patentreply.ai and the PatentReply record service that AI clients such as Claude and ChatGPT connect to. Most of what PatentReply works with is the public USPTO record; the rest is your account and the work you do on a matter, and that is what this policy is mostly about.
Effective October 2, 2026
What we hold, and for how long
Every kind of record PatentReply keeps about you is listed here, with its retention period. Deleted data can remain in encrypted database backups until those backups roll off, after which it is gone.
Your account
Email address, name, a password (stored hashed by our sign-in provider) or your Google sign-in, plan, and when you accepted the terms.
While your account is open; deleted within 30 days of a deletion request.
Billing
Stripe customer and subscription identifiers, plan, and the matters and charges billed. Card numbers are entered at Stripe and never reach us.
Kept after an account closes for as long as tax and accounting records require.
Your work on a matter
Documents you upload (a filed amendment, a non-patent reference, a draft response for checking), text you add, response plans, proposed claim corrections, approvals, generated claim listings and response documents, and check reports.
While your account is open; deleted within 30 days of a request. Download links expire after 7 days.
Compiled public records
The published application, its USPTO file wrapper, and the references and authorities it cites. These include names that appear in the public record, such as inventors, examiners and the practitioners who signed papers.
Kept so a matter reopens without recompiling. They are public documents and are shared across accounts.
Connector activity
For each call your AI client makes: the tool, one option value, the application number, your account identifier and plan, the outcome, timing and size, and the client’s name and version. Never what the tool was asked or what it answered.
13 months.
Product analytics
Pages viewed (address without its query string), product events such as a lookup or a sign-up, and, once you sign in, your account identifier and email. Pages that show matter content are excluded from automatic capture. Sessions are not recorded.
12 months, in PostHog.
Page-visit log
Path, referring page, browser user agent and time for public pages. No IP address.
13 months.
Free-tool requests
The patent or application number looked up, the result, your account identifier if signed in, and your IP address, which enforces the daily free-use limit.
IP address erased after 30 days; the request record is deleted after 13 months.
Service logs
Our hosts log each request to the record service and the site: the tool or page, your account identifier, the outcome and timing. Not the content of the request.
For our hosting providers’ log-retention period, then discarded.
Support
What you write to us, by email or through the support form.
While your account is open.
When you connect an AI client
When your AI client (Claude, ChatGPT, etc.) interacts with PatentReply, PatentReply may receive tool request messages and your sign-in token. We do not receive, and never ask for, your conversation history, the client’s memory, or your files in that client.
A tool request can carry text you wrote, such as an entry in a response plan, a reference you paste, or a search for MPEP sections. That text is handled as your work on the matter (above).
What we return to your AI client is data about the file wrapper you are interacting with. This information enters your conversation with your AI client provider, and what the provider keeps is governed by your agreement with them, not by this policy. For client matters, we recommend that you use a business or zero-retention plan with your AI provider (which can typically be configured in your AI client).
What our own processing sends to models
PatentReply also calls models itself (purely to do deterministic identification and indexing of a patent file wrapper; we do not analyze or perform interpretation of a file wrapper), through the companies listed below and on their API terms, which do not allow training on what is sent:
- At compile, public file-wrapper text is sent to Anthropic to identify the references an office action cites, and to a citation-reading model (TypeSafe, through Vercel AI Gateway, with zero data retention requested) to locate the passages the examiner relies on.
- PDFs are converted to text by Datalab. That includes wrapper papers and any non-patent reference you upload.
We do not train models on your matters, uploads, plans or approvals, and we do not use them to improve PatentReply for anyone else.
How we use your data
To run the service: compile and serve records, carry out the tools you or your AI client call, hold approvals, and generate documents. To sign you in and keep your account and matters separate from everyone else’s. To bill. To answer you when you write. To find and fix failures, and to see which features are used so we can improve them. To keep the service secure and within its free-use limits. To meet legal obligations.
We do not sell personal information, do not share it for advertising, and do not show ads.
Who processes your data
These companies process data on our behalf, under contract, for the purposes above:
- Supabase
- Database, file storage, and sign-in (the OAuth issuer your AI client signs in through)
- Vercel
- Hosts patentreply.ai; cookieless page-performance analytics; AI Gateway, which routes model requests
- Railway
- Hosts the record service your AI connects to (mcp.patentreply.ai) and its logs
- Cloudflare
- Edge network and TLS in front of the site
- Anthropic
- Identifies and indexes the cited references in a file wrapper at compile, on API terms
- TypeSafe
- Citation-reading model, reached through Vercel AI Gateway with zero data retention requested
- Datalab
- Converts PDFs to text: USPTO file-wrapper papers, and non-patent references you upload
- Stripe
- Payments and invoices; card details never reach PatentReply
- Resend
- Account and billing email
- PostHog
- Product analytics: usage events with account and application numbers, never matter content; no session recordings
- WorkOS
- Sign-in for firm service accounts only
To fetch public documents and authorities, requests also reach these services, which are not processing for us:
- USPTO Open Data Portal
- Receives application numbers when a file wrapper is compiled or refreshed
- Google Patents
- Receives patent and publication numbers when a cited reference is fetched
- mpep.io
- Receives the MPEP and case-law search text your AI sends to the authority tools
If you sign in with Google, Google tells us your name and email. Beyond these, we disclose personal information only when the law requires it, to protect the rights or safety of users or others, or to a successor if the business is sold or merged, which would remain bound by this policy.
Your choices and rights
You can ask to see the personal information we hold about you, correct it, receive a copy, or have it deleted. A matter’s compiled record can be exported at any time with the export tool; for a copy of everything else we hold about you, ask. To make a request, write to [email protected] from your account email; we answer within 30 days. Deleting your account deletes your private data on the same terms; compiled public records and billing records are kept as described above.
To cut off an AI client, remove the PatentReply connector in that client, or use Sign out everywhere on your Settings page to revoke every client at once.
California residents have the rights above under California law. We do not sell or share personal information as that law defines those terms, and we will not treat you differently for exercising a right. Where the law of the place you live gives you further rights, write to us and we will honor them.
Security and location
Data is encrypted in transit and, by the providers that store it, at rest. Each account’s work on a matter is kept apart from every other account’s, and a test of that separation runs on every code change. The Security page describes the controls, including what has not yet been audited. PatentReply’s databases and file storage are in the United States; the providers above may process data where they operate.
Children
PatentReply is a professional service for patent practitioners and is not directed to children. We do not knowingly collect personal information from anyone under 16; if we learn we have, we delete it.
Changes
When this policy changes, the effective date at the top changes with it. If a change gives us more room to use your information, we email account holders before it takes effect.
Contact
Patent Agentic LLC · [email protected]