AI on your server. Your data goes nowhere.
We deploy a private AI assistant — LLaMA 3 or Mistral — on a server you control. Your team gets a capable AI tool trained on your internal documents. Nothing leaves your infrastructure.
Who it works for
- Law firms, accounting practices, and financial advisors handling confidential client data
- AEC firms with proprietary project files, engineering drawings, and client contracts
- Medical and allied health practices with patient records and clinical documentation
- Any SMB that uses AI but is uncomfortable sending internal documents to a third-party API
Who it does not work for
- Businesses with no server budget — minimum viable hardware costs money to run
- Teams that have no sensitive data and are happy using ChatGPT or Claude directly
- Companies with no internal documentation to train on — the assistant needs a knowledge base to be useful
✦ What's included
Everything in the setup
Data stays on your server
Queries, documents, and conversation history never leave your infrastructure. No data passes through OpenAI, Anthropic, or any third-party API.
Trained on your documents
SOPs, project files, pricing sheets, contracts, client records — the assistant answers questions based on your actual internal knowledge, not general internet knowledge.
Team access with permissions
Role-based access control so each team member sees only what they should. Engineering sees project files; finance sees financial records; management sees everything.
Runs on your hardware
Deployed on a VPS you control or a physical server on your network. You own the infrastructure. We handle the configuration.
Connects to your tools
We expose an internal API endpoint so the assistant can be accessed from Slack, your project management tool, or any internal application your team already uses.
Stays current
When better open-weight models release — and they release frequently — we evaluate and update. Your assistant improves without you managing it.
✦ How it works
From audit to live assistant
Audit
We review what you want the assistant to do, what documents it should know, who needs access, and what hardware you have available or are willing to provision.
Deploy
We configure the server, install and test the model, and build the private document index. This takes one to two weeks depending on document volume.
Integrate
We connect the assistant to your team via a simple interface and, where applicable, to your existing tools through the internal API endpoint.
Hand over
We document how to use it, train your team, and monitor the first month. After that we stay on retainer for updates and support.
✦ Pricing
Private AI pricing
One setup fee to deploy, configure, and hand over the assistant. Then $197/month to keep it running, updated, and maintained.
Setup
Server configuration, model deployment, and knowledge base integration
- Server or VPS audit and specification
- LLaMA 3 or Mistral model deployment
- Private document index (RAG setup)
- Team access configuration and permissions
- Internal API endpoint for tool integrations
- Staff onboarding and usage documentation
- 30-day post-launch monitoring
Support
Monitoring, updates, and knowledge base maintenance
Covers model updates, server monitoring, and document index refreshes.
- Monthly model update review
- Document index refresh as content changes
- Server performance monitoring
- Access control updates
- Priority support for downtime or errors
✦ FAQs
Common questions about private AI
✦ Get started
Book a free assessment
We will review your infrastructure, use case, and documents — and tell you exactly what a private deployment would look like for your business.