Stack
OpenAI · Anthropic · Gemini
Frontier models, one gateway, observable from day one.
We route every model call — OpenAI, Anthropic, Gemini and open-weight models — through a single gateway. That gives you per-feature cost limits, fallbacks when a provider goes down, full logs of every prompt and the ability to swap models without touching product code.
Why we use it
- Swap models per feature without redeploying the app
- Automatic failover when a provider has an outage
- Per-feature cost caps so a runaway loop can't ruin your month
- Full prompt and response logging for debugging and evals
Where it shows up in our builds
- Support copilots, summarisers and inbox triage
- RAG over your docs, tickets and database
- Lead-magnet generators and AI-powered onboarding flows
Services that lean on it
Also in the stack
Lovable
AI-native build environment
React + TanStack Start
App framework & SSR
TypeScript
Across the whole stack
Tailwind CSS
Design tokens, never inline colours
Supabase
Postgres, auth, storage, edge functions
Cloudflare Workers
Where everything runs
Stripe
Payments and subscriptions
Resend
Transactional email
Framer Motion
Soft, considered motion