The full patent record,
wherever you work.
Advanced patent search in your browser or inside Claude and ChatGPT. Verbatim text cited to the column and line, the claims as they stand today, file histories, and court decisions, all in one record.
One place. Everything.
Today you’re switching between Google Patents or USPTO for search, Patent Center for File Wrappers, Global Dossier for foreign continuity, and PTAB E2E for proceedings. Patent Bear pulls it all into one record, puts it inside your AI, and adds the workflow, annotation, and analysis layers the rest don’t have.
BearGoogle
PatentsPatent
Public
SearchPatent
CenterGlobal
DossierPTAB
E2EGoogle
Scholar
Feature comparison reflects publicly documented capabilities of each tool as of October 2026. Google Scholar includes patents, but it is a scholarly-search product rather than a dedicated patent-search workspace; Google Patents is Google’s patent-specific search surface. Sources: Patent Public Search, Patent Center, Global Dossier, PTAB E2E / P-TACTS, Google Patents, Google Scholar.
Give Claude and ChatGPT the full patent record.
Patent Bear connects to the AI your team already uses. Ask a patent question in plain English and your assistant works from the record itself: verbatim text, the operative claims, the file history, the court decisions, all cited to the source.
Every quote cited to the column and line
Your assistant reads the issued text, not its memory of it. Every passage comes back verbatim with a column and line cite taken from the printed patent, and paragraph numbers for publications and European documents.
Each cite links to the passage, highlighted in the reader with the PDF alongside, so checking a quote before a brief goes out takes seconds.
Claims as they stand today, not as they issued
Claims come back in their operative form after any reexamination or PTAB certificate, with cancelled claims marked. Term, expiration and maintenance-fee status come with them, and current ownership from the assignment record.
The fact a model on its own most often misses is the one that matters most: whether the claim is still alive.
Office actions and responses, read in full
Ask about an application and your assistant opens the file wrapper itself: the rejection, the references, and how the applicant answered, inside the conversation.
No downloading documents from Patent Center one at a time to upload them back into a chat.
What the courts held, patent by patent
Nearly 60,000 district court, Federal Circuit and Supreme Court decisions, searchable by patent, party, judge, issue and doctrine, with the ruling, the ground and the claims involved.
Every holding comes with the opinion text behind it and a cite to the page, so the summary can be checked against the court's own words.
Search like a searcher, in a sentence
Owner and applicant, inventor, CPC exact or with subgroups, filing, issue and publication dates, claim text and kind, across US and European records, plus semantic similar-patent search and bulk sweeps of up to 1,000 results.
Totals are real and owner names are merged across spellings, so "who is filing the most" gets a straight answer.
Zero data retention at every AI provider
Every AI provider Patent Bear uses is bound not to keep what it is sent, and nothing is used to train a model. Each user signs in with their own account and sees only their own work.
Standard MCP with OAuth. Claude, Claude for Work, ChatGPT, Gemini and Codex, in about two minutes per user.
Patents, publications, and applications, one surface.
Search by number, keyword, assignee, or inventor. Jump straight to claims, figures, prosecution history, file wrappers, family trees, and the court decisions on each patent. Always up to date, from the USPTO, the EPO and the courts.
Never lose a patent you looked at
Remember that patent you found last Tuesday? The one with the interesting claim structure? No more wasting time endlessly looking through your browser history.
Patent Bear automatically saves every patent you visit with its title, so you can find it instantly. Review your search history anytime.
Every Office Action, every applicant remark, in one place
Open any application to see its complete file wrapper (Office Actions, applicant remarks, amendments, Notices of Allowance, examiner interview summaries) alongside the spec.
No more bouncing to Patent Center. The full prosecution history loads inline with the application data, with continuity data and the prosecution timeline visible at a glance.
Every relative of a patent in one place
Continuations, divisionals, CIPs, provisionals, and foreign counterparts, all in a single visual tree. No more piecing the family together across Global Dossier, Patent Center, and PAIR.
Click any node to jump to that patent or application. The tree shows priority dates and status at a glance, so you can see where a claim line started and how it branched.
Court decisions, searchable and readable
Search decisions by party, judge, court, patent, issue or words in the opinion, then read them in the same reader you use for patents, with rulings and holdings laid out alongside.
Open any patent and the decisions that ruled on it are already there: invalid, not infringed, or construed, with the ground and the claims involved.
Examiners and judges, by the numbers
See how an examiner has handled applications in an art unit, or how a judge has ruled on an issue, built from the office actions and decisions themselves.
Useful for planning a response, choosing a venue, or setting a client's expectations before the next filing.
Built for how patent lawyers actually work.
Color-coded highlights. Per-claim term tracking. Figure annotations that stick. Claim charts without the spreadsheet. Everything persists across sessions and devices.
Highlight, annotate, and track claim terms
Select any text and highlight it in 10 colors. Add notes to any highlight. Everything persists across sessions and devices.
Select a term in a claim and it's automatically found in every other claim. Navigate between occurrences with a single click. Essential for claim construction and prior art mapping.
Jump to any section in one click
Every patent has a built-in table of contents. Jump straight to claims, figures, or any section of the specification.
The specification headings are pulled directly from the patent. Expand to see every section. No more endless scrolling.
Break claims into elements, map your analysis
Automatically create claim charts. Patent Bear parses each claim into its preamble and individual elements: [1Pre], [1a], [1b], [1c]. Add notes alongside each element in a two-column chart.
Edit element text to refine the breakdown. Export the finished chart to Word for your brief or client report. Duplicate charts for the same patent.
AI is only useful when it can work from the record.
Ursa is not a generic prompt box bolted onto a patent site. Patent Bear structures the claims, specification, figures, prosecution history, families, and citations first, then gives AI the tools to reason against that record, show its work, and turn plain-English research goals into reviewable search and analysis.
LLMs are powerful alone. Tools unlock another level.
A raw LLM can draft fluent text. Patent Bear gives it the instruments needed for serious patent work: structured claims, specifications, figures, prosecution history, family data, citations, search tools, and work-product workflows.
That tool layer is the difference between a model talking from memory and AI working from evidence. Patent Bear gives LLMs the record and the tools to search, compare, map, cite, and produce analysis a patent professional can verify.
Ask Ursa anything about any patent
Chat with Ursa, Patent Bear's AI patent assistant, against every word of the patent you're reading. Get plain-English answers about claim scope, prior art implications, and technical details, cited back to the source.
"What problem does this patent solve?" "Summarize claim 1." "Which dependent claims add hardware limitations?" Every answer links to the specific claim, paragraph, or figure it came from.
Plan research before you know the perfect query
Use Ask Ursa for patent research questions, search planning, invention ideas, and analysis strategy that span more than one patent record.
Ask in plain English, then move from the answer into searches, patent-specific chat, or an analysis when the question turns into work product.
Turn plain English into advanced search fields
Describe the search you want, and Ursa Suggestions structures it into reviewable advanced-search filters before you run anything.
Ursa can suggest keywords, assignees, inventors, owners, CPC classes, dates, document groups, and match fields. It fills the form; you stay in control of the search.
Click a citation, jump to the source
Every AI answer contains inline citations. Click “Claim 1” in the response and the patent scrolls to claim 1. Click “¶ 42” and it jumps to that paragraph.
Less guessing, fewer unsupported summaries, and a clear way to verify whether the AI actually read the patent. The citation is the proof.
The analyses you'd pay an associate for. In minutes.
Prior-art searches, invention screens, §112 disclosure checks, freedom-to-operate scans: the work that used to take a week of an attorney's time, started in a sentence from the site or from your AI. Every output is grounded in the record and cited back.
§102 / §103 invalidity, with claim charts
Prior Art analysis does not hand a prompt to a third-party patent-search API and hope for the best. Patent Bear controls its own patent-search database of 20M+ patents and publications, so it can search aggressively across the corpus.
A single run can fan out into thousands of search operations: element-by-element queries, alternate phrasings, citation and family exclusions, CPC-focused passes, and candidate re-ranking. That is how the analysis finds meaningful art instead of stopping at the first obvious references.
Search from an invention disclosure before you draft
Describe the invention in ordinary technical language. Patent Bear turns the disclosure into structured search targets and looks for nearby patents, publications, and technical literature.
Use it before claims exist: screen a concept, pressure-test novelty, and surface references that should shape drafting strategy before the application is written.
§112 support, claim by claim
Checks each claim against the specification for written description, enablement and definiteness, with the supporting passages cited.
Useful before filing, in prosecution, and when testing a granted claim's support in a validity fight.
Freedom-to-operate, starting from the product
Describe a product and the analysis scans for live claims that could block it, with the closest claims and the passages that matter.
A first pass that shows where to look hard, not a legal opinion.