Know what is really in a PDF before an AI reads it
Nelix finds hidden text, suspicious formatting and prompt-injection signals in PDF documents, and shows you the evidence. Made by Clouxart Ltd in Cambridge.
What we are for
To let you trust the documents your AI reads. A finding is only useful with its evidence, so every result shows where in the file it came from and why it matters.
How we work
Four things you can hold us to
Evidence, not verdicts
Every finding points to the place in the file it came from, so you can check it yourself.
Plain findings
Hidden text, text outside the page, fonts too small to read, encoding oddities: named as what they are.
Open core
The scanner is open source as doc-sherlock. The service adds accounts, an API and history.
A person answers
Nelix is made by two people. Write to us and one of us replies.
How Nelix came about
Nelix started in April 2025 with a simple observation: a PDF can look harmless to a person and still carry instructions that steer an AI. White text on a white page, text placed outside the page, fonts too small to see. We wrote a scanner to find those signals and published it as open source, doc-sherlock.
The service grew from the scanner. Upload a document, or send it through the API, and Nelix reports what is hidden in it, with the evidence for every finding, so you can decide what to trust before the document reaches your AI. It is a small product made by Clouxart Ltd in Cambridge, and the people who build it are the people who answer your email.
A scanner for hidden instructions in PDFs, and the first service that runs it.
The scanner published on GitHub as doc-sherlock.
Anyone can sign up, scan a document and pay for more.
Founders
Onuralp Taner
Founder & Principal Engineer, Clouxart Ltd. Builds and runs Nelix.
Paolo Di Prodi, PhD
Ex-Microsoft and ex-Fortinet engineer, working on AI safety and security.
Try it on one of your own PDFs
A free account is enough to see what Nelix finds.