◯ AI

AI you actually own.

Most "AI hosting" sends your prompts to OpenAI and calls it private. We run models on our own hardware by default, and let Business customers route to their own API keys when they want a specific provider.

◯ Built-in

Three AI features. No setup.

◯ Assistant
Chat assistant

Available on every page. Ask about plans, billing, domains, or how to use any Vinmur feature. Streams responses word-by-word.

◯ Builder
AI site builder

Describe your business, get a complete one-page site — copy, layout, fonts, all of it. Edit it section by section in plain English. No drag-and-drop required.

◯ Suggester
Domain suggester

Describe what you're building, get 6 brandable domain ideas with real-time availability checks across .com, .co, and .ai.

◍ The privacy default

Your prompts stay on our hardware.

By default, Vinmur's AI runs on local infrastructure — an NVIDIA Blackwell GB10 in the UK running open models via Ollama. Your chat messages and domain queries never touch OpenAI, Anthropic, or any third-party API.

That's a real privacy guarantee, not the kind where “private” means “we proxy to OpenAI and promise we won't look.” The model runs on our machine. Your data doesn't leave.

Chat + Domain suggest
Local — Qwen 2.5 32B

Open-weights model running on the Vinmur GPU. No third-party API calls for these features.

Site builder
Groq — Llama 3.3 70B

The site builder uses Groq for speed (sub-second time to first token). On the BYOM roadmap to make this fully local too.

◍ Business plan

Bring your own AI provider.

Want to use GPT-4o for chat, Claude for the site builder, and a custom Fireworks endpoint for domain suggestions? On the Business plan, you do.

Add your OpenAI, Anthropic, Groq, or any OpenAI-compatible API key in the dashboard. Toggle which feature uses it. Your prompts then run on YOUR provider, billed to YOUR account. Switch back to Vinmur defaults anytime.

Step 1
Add your key

In the dashboard, paste your API key. We test it before saving and encrypt it at rest with AES-256-GCM.

Step 2
Route per feature

Toggle which AI feature uses this key — chat, site builder, domain suggest, or any combination.

Step 3
Use as normal

Vinmur features work exactly the same. The difference is your prompts go to your provider, your invoice.

OpenAI
gpt-4o, o1, ...
Anthropic
Claude Sonnet, Opus, Haiku
Groq
Llama, Mixtral, Qwen
Custom
Fireworks, Together, OpenRouter, your own Ollama

Private by default, flexible by design.

Pro for managed Postgres + full AI features. Business for dedicated containers + private model hosting + BYOM.