Example output. Your code is analyzed locally and never stored raw.
Powered by Official USPTO Data
Run our local extractor in your own editor. It reads your repo on your machine, pulls out only the abstract ideas, and sends us nothing but a few plain-English invention descriptions. Your source never leaves your laptop.
The local skill analyzes your code in your own session. Only abstracted invention descriptions are sent, never source.
On the server-scan option, we redact API keys, tokens, and private keys before any AI step, and never read .env or key files.
Code is scanned in memory and discarded. We keep only the derived analysis. Embeddings run on our own infrastructure.
Read the details in our Privacy Policy and confidentiality terms.
Link GitHub, Jira, and your design docs. Or run the local skill so nothing but the ideas ever leaves your machine. Patentable reads what your team actually built.
Inventions surface on their own, scored for novelty and Alice survivability and grounded against 44M+ patent claims. You get an inbox of what is worth protecting, ranked T1 to T3.
Generate an attorney-ready provisional draft, get referred to a patent attorney to review and file, and a weekly update keeps watching as your code evolves.
Want to search patents directly instead? Browse the 44M+-claim patent database.
Point it at your work. It finds the IP, scores it, and gets it ready for an attorney.
Run the extractor in your own Claude Code. It reads your repo locally and sends us only the abstracted inventions. We never see a line of your source.
Prefer zero setup? Connect a repo plus your design docs and we scan it server-side, secrets stripped, code never stored raw.
Each invention is scored for novelty and Alice survivability and grounded against 44M+ patent claims, then ranked T1 to T3.
Generate an attorney-ready provisional draft, included with Pro. We refer you to a patent attorney to review and file.
Every claim is translated from legalese to developer language, so when we ground your inventions against the prior art, we catch matches no matter how the lawyers worded them.
Patent lawyers and developers describe the same concepts differently. We index every claim twice, in legalese and in plain English, so your inventions match the prior art no matter how it was worded.
We find both. Every time.
The local extractor reads your code on your machine and sends only an abstract invention description. There is no field for source code, so it cannot be sent even by accident.
# Extracted locally, in your own editor. We never receive your code.
# Only this abstraction is uploaded:
{
"title": "Patent-domain embeddings for cross-source dedup",
"description": "Deduplicates discovered inventions across repos and
docs by comparing patent-domain embeddings rather than text.",
"technical_approach": "Embed each candidate, cluster by cosine
similarity, keep the highest-novelty representative.",
"problem_solved": "Same invention surfacing from multiple sources"
}
# No source code. No file contents. No snippets. By design.See what is patentable in one codebase for free. Add seats when your team is ready.
Provisional drafts included, not billed per document
Get qualified leads from developers who need help filing. We connect inventors with attorneys.
Join our attorney network →