Skip to content
Tell me your problem Tell me

Private AI

A language model running on a machine you own, with access to your documents and your systems. I install it, connect it and teach you to use it. Your data stays on your network and there is no per-query bill.

Tell me your problem Message me on WhatsApp

Who it's for

At home

  • You want a voice assistant that runs the house and understands you without sending audio anywhere.
  • You dictate notes and want them to land, tidy, in your own notes.
  • You like the idea of ChatGPT, just not the idea of it reading your life.

In your business

  • Your team asks the same question twenty times a day and the answer is in a PDF nobody can find.
  • You answer the sales WhatsApp by hand, including at eleven at night.
  • You want AI over contracts, quotes and customer data without uploading them to a third-party service.

What's included

  1. A prepared server: Proxmox and Ollama on your mini PC, workstation or server.
  2. An open model sized to your machine (8B to 70B depending on the package).
  3. Your documents connected: the assistant answers citing your PDFs, spreadsheets and manuals.
  4. Krynth AI: the chat interface with users and permissions, built on open-source software (Open WebUI or LibreChat), so the whole team signs in with their own account.
  5. Secure remote access over VPN and automatic backups.
  6. Integrations by package: WhatsApp, CRM or ERP, voice.
  7. Training and an adjustment period after delivery.

What I leave out

  • Training or fine-tuning models from scratch: for what I sell, connecting your data is enough.
  • Consulting without implementation: if I don't build it, I don't bill it.
  • The hardware: I tell you what to buy and, if you prefer, I source it at cost plus 10%.

Packages in this line

Prices in Mexican pesos (MXN), before tax. Hardware is extra. The final price comes out of the diagnostic. US dollar figures are approximate, at the exchange rate of October 20, 2026.

AI at home

From $15,000 MXN≈ US$810

Includes

  • Mini PC with Proxmox and a local model (Ollama)
  • Local voice assistant tied to Home Assistant
  • Voice notes that land in Obsidian
  • VPN to get in from outside
  • Two hours of training

Not included

  • The mini PC ($11,000 to $14,000 MXN)
Time
2 to 3 weeks
Payment
50% to start, 50% on delivery
Ask for this package

Private AI S

From $45,000 MXN≈ US$2,400

Includes

  • Mini PC with Proxmox and Ollama (8B to 20B models)
  • Your documents connected, up to about 5,000 pages
  • Krynth AI, the interface with users and permissions
  • Backups and VPN
  • Four hours of training
  • 30 days of adjustments after delivery

Not included

  • Hardware ($15,000 to $25,000 MXN)
Time
3 to 4 weeks
Payment
50% to start, 50% on delivery
Ask for this package

Private AI M

From $95,000 MXN≈ US$5,100

Includes

  • Everything in S, on a GPU workstation (10 to 50 users)
  • 30B to 70B model
  • One integration: CRM, ERP or WhatsApp
  • Hybrid: local, with an external API only for peaks
  • Monitoring

Not included

  • Hardware ($90,000 to $120,000 MXN)
Time
5 to 8 weeks
Payment
40% to start, 30% at the demo, 30% on delivery
Ask for this package

Private AI L

From $180,000 MXN≈ US$9,700

Quoted to scope

Includes

  • Multi-GPU server
  • Single sign-on (SSO)
  • Several integrations
  • Service level agreement
  • Operations documentation

Not included

  • Hardware (from $250,000 MXN)
Time
2 to 4 months
Payment
30% to start, 30% at the demo, 40% on delivery
Ask for this package

WhatsApp bot

From $25,000 MXN≈ US$1,400

Includes

  • Conversation flow designed with you
  • Answers from your documents
  • Handoff to a person when needed
  • Panel to see the conversations

Not included

  • WhatsApp API usage (Meta charges per message: US$0.008 utility, US$0.0436 marketing)
Time
2 to 4 weeks
Payment
50% to start, 50% on delivery
Ask for this package

Voice agent

From $35,000 MXN≈ US$1,900

Includes

  • Call flow designed with you
  • Answers from your documents
  • Handoff to a person when needed
  • Call panel

Not included

  • Voice usage (ElevenLabs charges per minute)
Time
2 to 4 weeks
Payment
50% to start, 50% on delivery
Ask for this package

Frequently asked questions

Why local AI and not ChatGPT?

For anything that isn't sensitive, use ChatGPT: it's good and cheap. Your contracts, your customers and your costs shouldn't leave your office, and a local model has no per-query bill.

What happens to my data in a private AI?

It stays on your machine. The model runs on your server, your documents are indexed there and queries never leave your network. In hybrid packages, only what you mark can go to an external API during usage peaks, and we put it in writing. If you'll load customers' personal data, I include a confidentiality agreement and help you reflect it in your privacy notice.

Does the price include the hardware?

No. Each package states what the equipment it needs costs, so you can buy it yourself from my list. If you'd rather I source it, I charge cost plus 10%, no hidden markup.

What if the hardware fails?

You have the manufacturer's warranty and a daily backup of your data. With a monthly plan, I restore it on another machine.

What if you disappear one day?

Everything stays on your hardware and in your accounts, documented, with the code in your name once the project is paid. Any developer can pick it up without asking me.

Who maintains it afterwards?

What I deliver has a 30-day warranty and a free check-up on day 30. After that you can carry on alone (everything is documented), buy a block of hours, or take a monthly plan with monitoring, updates and a defined response time, from $800 MXN per month for a home and $4,500 MXN for a business.

Tell me what you want to fix.

Twenty minutes on a call, free. You leave with an answer: which package fits, or whether a diagnostic comes first.

Tell me your problem