About Kognix
Kognix indexes only what you give it, retrieves only from what you uploaded, and cites every claim back to the original passage. No guessing. No hallucinations.

The problem
General-purpose AI tools are trained on the entire internet. They're great for broad questions — but the moment you ask about your professor's lecture slides or a specific company report, they guess to fill the gaps. Those guesses sound confident, but they're not grounded in your source material.
Kognix solves that. It indexes only what you give it, retrieves only from what you uploaded, and cites every claim back to the original passage. If the answer isn't in your documents, it says so.
Without Kognix
What does my professor say about flash floods?
Flash floods are rapid floods caused by heavy rainfall...
⚠ From the internet, not your notes
With Kognix
What does my professor say about flash floods?
According to your AHS 106 notes, flash floods result from intense rainfall over a short period...
✓ From your uploaded lecture notes
Values
Four principles guide every decision — from how data is stored to how answers are presented.
Grounded by default
Every answer cites the exact passage it came from. If the answer isn't in your documents, Kognix says so.
Your data stays yours
Documents are stored per-user. Kognix never trains on your uploads or shares them across accounts.
Open source
The full stack is MIT licensed and available on GitHub. Audit the retrieval path or self-host entirely.
Simple and focused
Upload, ask, get cited answers. No bloat, no dashboards you'll never use, no features that get in the way.
Open source
Self-hosting coming soonKognix is open-source under the MIT licence. The full stack — Next.js frontend, .NET API, Pinecone vector store, PostgreSQL — is available to fork, audit, and extend.
Get started
Start with your documents and see the difference grounded AI makes. Free to try — no credit card required.