What is odnoga?

odnoga is an AI gateway and control plane for multi-tenant SaaS. One OpenAI-compatible endpoint, every major model vendor, and the accounting, governance and prompt tooling a product with its own customers needs.

What is odnoga?
odnoga is an AI gateway and control plane for multi-tenant SaaS. It exposes one OpenAI-compatible API at https://api.odnoga.com/v1 that routes requests to OpenAI, Anthropic, Google Gemini, Mistral, Groq, xAI and other vendors, while metering every request per tenant and per end user.
Who is it for?
Teams shipping AI features inside a product that has customers of its own: SaaS vendors, agencies and platforms that must attribute AI cost to a tenant, cap spend, and prove where data was processed.
How does odnoga work with Supabase?
You call odnoga from a Supabase Edge Function with a plain HTTP request — no package to install. Forward the end_user_id you read from the Supabase JWT and every call is metered, budgeted and traced per user. odnoga itself runs on Supabase — Postgres, Edge Functions, RLS and pg_cron.
What makes it different from a proxy?
A proxy forwards traffic. odnoga adds a versioned prompt registry, routing policies with fallbacks, per-request cost ledger, tenant and workspace budgets with approval gates, data-residency enforcement with vendor evidence, and an MCP server so Claude, Cursor and Lovable can read and change that configuration directly.
Where is data processed?
You choose a default region per tenant or workspace (EU or US). In strict mode only models with a matching regional endpoint are callable; requests to non-compliant vendors are rejected. Each request stores the requested region, whether it was actually honored, and the vendor endpoint used.
How is odnoga priced?
You pay vendor cost plus a transparent platform margin, shown per model in the catalog before you send a request. Regional processing adds the vendor residency uplift only when traffic is genuinely routed to that region.