Private AI assistantsAnswers from your own handbook, with citations.
An AI that knows your company and never phones home.
In production, running our own company. We can screen-share it today.
See it step by step
Swipe for the next step
The problem
The hard part isn't the chatbot. It's grounding.
A general chatbot doesn't know your policies, and sending company documents to an outside service means your data leaves your control. Teams need an assistant that answers from their own material and shows where each answer came from.
What we provide
A private assistant, running on your infrastructure.
Self-hosted model
Qwen2.5-14B served with vLLM on servers you control — both Apache-2.0 licensed.
Wired into team chat
Ask questions where your team already works.
Answers from your documents
Grounded in your actual handbook and policies, with source citations.
Confidential fields stripped
Salary, bank details and keys are removed before the model ever sees them.
Built on open source
Qwen2.5
vLLM
What it replaces
AI help, without sending your data to someone else's cloud.
Instead of pasting company material into a public AI service, your team asks an assistant that runs inside your own environment.
- Instead ofPublic AI chat servicesYou runA model on your own servers
- Instead ofSearching shared drives by handYou runCited answers in chat
What's included
Set up, grounded and looked after.
- Model deployed on infrastructure you own
- Connected to your team chat
- Your documents indexed for cited answers
- Confidential-field filtering before the model
- Run, patched and monitored by us