Input
Supported single-turn text
POST /v1/chat/completions and GET /v1/models are supported within the published compatibility boundaries.
TOKENFLOW / PRODUCT GUIDE
What you can build with TokenFlow today.
A focused developer gateway for supported, single-turn AI text requests.
TokenFlow provides one controlled entry point for selected OpenAI, Anthropic and Google Gemini request formats. You retain your own provider credentials and relationship; TokenFlow provides its own gateway credential and request-level operational visibility.
Open developer portalLaunch contract
We describe the release as it operates, not as a promise about every future capability.
Providers
Bring your own keys
Use your TokenFlow tenant key plus the provider key you hold. Provider billing, availability and rate limits remain with that provider.
Control
Explicit routing only
A cross-provider route requires explicit authorisation and a suitable destination credential. Requests are not silently transformed.
Included at launch
- Supported OpenAI, Anthropic and Google Gemini API shapes.
- Model-aware routing decisions within a selected provider.
- Bounded admission, queueing and retry behaviour.
- Tenant-scoped usage and operational telemetry.
Outside the default release
- Multi-turn histories, tool calls and multimodal requests.
- Universal IDE, desktop chat or agent-framework compatibility.
- Guaranteed provider availability, cost savings, speed or 429 avoidance.
- Pooling customer credentials or treating provider quota as transferable capacity.
Why the boundaries matter
Supported request shapes are handled explicitly; unsupported shapes are rejected instead of being silently reinterpreted. This gives developers a reliable base for evaluation and gives platform teams a clear operational contract.