TensorLoopDocumentation
Build with TensorLoop, fast.
One scoped key, an OpenAI-compatible endpoint, and a fleet of open-source models behind it. These docs cover everything from your first chat completion to per-key budgets, streaming, and tool calls.
Quickstart · ~ 5 minutes
3 steps- 01
Sign up
Create a free account. You start on the free plan with two keys and gpt-4o-mini.
- 02
Mint an API key
Set a budget, RPM, and model allowlist. The key appears once — save it to a password manager.
- 03
Send a request
Point any OpenAI client at our base URL. Watch tokens and cost land in your dashboard within seconds.
Looking for something specific?Search every page, parameter, error code, and example.
⌘KExplore the docs
StartQuickstartSign up, mint a key, and make your first chat completion in under three minutes.FoundationsConceptsThe mental model — accounts, plans, keys, budgets, and the two surfaces.AuthAuthenticationHow keys are scoped, minted, and revoked. Token format and rotation.ReferenceChat completionsPOST /v1/chat/completions — parameters, message format, tool calling, vision.ReferenceStreamingSSE wire format, finish reasons, reconnection guidance.ReferenceModelsGET /v1/models — list models available to a bearer key.LimitsRate limits & budgetsPer-key RPM caps, dollar budgets, and what happens when you exceed either.By taskExamplesDrop-in snippets for curl, Python, JavaScript, streaming, and tool calling.When it breaksTroubleshootingWhy your key is 401-ing, why spend looks wrong, why streams disconnect.ReferenceErrorsEvery code TensorLoop and its upstream proxy can return, with fixes.