The control plane for model access.
GotoAI sits between your application and many model providers. It speaks the OpenAI API, routes each request, gates risk, settles a ledger, and gives you a console to run it all.
Five layers, one request path
OpenAI-compatible by design
Chat completions, streaming and a models endpoint — the same shapes your code already uses.
- Drop-in for the OpenAI SDK
- SSE streaming, chunk-by-chunk
GET /v1/modelsreflects your catalog
GET {{api_base}}/models
POST {{api_base}}/chat/completions
# Authorization: Bearer $GOTOAI_API_KEYRoute across providers, fail over safely
The same model family on different providers is a separate endpoint with its own price, latency and policy. The router picks the best by mode and health.
- Modes: cheap, quality, latency, privacy, code
- Health scoring + circuit breakers
- Same-tier fallback, including pre-first-token for streams
| Mode | Picks |
|---|---|
| cheap | lowest output price |
| quality | highest quality score |
| latency | fastest healthy provider |
| privacy | no-log providers only |
Settle every request
Each call is a reserved order, finalized by real tokens, recorded in an append-only ledger and reconciled for drift.
- provider cost · platform fee · user charge
- Idempotent, replay-safe money operations
- Automated reconciliation with alerting
reserve −$0.0009 (max possible) execute provider=openrouter finalize −$0.0000071 (real tokens) release +$0.0008929 (unused hold)
Operate the mainnet
A control panel for keys, the playground, usage, the ledger, supply management and risk — for users and platform admins.
- Create keys & debug in the playground
- Inspect usage and reconcile billing
- Reprice / disable providers, set kill switches
| Page | What it does |
|---|---|
| Playground | stream & debug calls |
| Billing | wallet, ledger, reconcile |
| Supply | endpoints, pricing, health |
| Risk | limits, kill switch |
Governance for teams
BYOK, audit export, no-log enforcement and org-scoped administration are behind the Enterprise stage gate during S1 beta.
- BYOK — gateway-fee billing Planned
- Audit export, metadata only Planned
- Org RBAC & SSO Planned
# Enterprise stage gate is closed during S1 beta
# Roadmap: POST {{api_base}}/enterprise/byok
# provider=openai, key encrypted at rest
# requests use your key → gateway fee onlySee it in your stack.
Create a key and route your first request through GotoAI.