Frequently asked questions

Flux is an LLM gateway. One API key gives you access to 30+ models behind a single OpenAI- and Anthropic-compatible endpoint. Smart routing right-sizes each prompt to a model that can handle it, so you stop paying top-tier rates for work a cheaper model does just as well. Billing is pay-as-you-go, and it's materially cheaper than calling the providers direct.

Sign up, then create a key from your dashboard. Keys are in the format sk-... and work as a normal bearer token. You use the same key for every endpoint and every model.

Point your client at https://api.fluxrouter.ai, pass your key as "Authorization: Bearer sk-...", and send a request to POST /v1/chat/completions with "model": "flux-auto". flux-auto lets Flux route the request for you. That's it. If you already have working OpenAI code, only the base URL and key change.

Use whichever you already have. The OpenAI-compatible endpoints are POST /v1/chat/completions, POST /v1/responses, and GET /v1/models. The Anthropic-compatible endpoint is POST /anthropic/v1/messages. Any SDK that talks to those shapes works once you set the base URL to https://api.fluxrouter.ai (or .../anthropic for the Anthropic path).

You pay per million tokens based on the routing lane your request lands in. There are no seats and no minimums on pay-as-you-go. You're charged for the input tokens (your prompt and system message) plus the output tokens the model returns.

Express Lane starts at $1 per 1M tokens, Daily Driver starts at $2 per 1M tokens, and Deep Thought starts at $4 per 1M tokens. Lighter prompts route to cheaper lanes; harder prompts that need more capability route to higher lanes. See the pricing page for the current per-lane input and output rates.

Yes. The Free plan gives you $1 of routing per month at no cost, with a hard stop at the cap so you never get a surprise bill. It's a real taste of the smart router, not a crippled demo. A card is required at signup as an anti-abuse measure and is not charged on the Free plan.

By usage. Every request is metered on the tokens it actually used at the routed lane's rate, and that's what you pay. Pay-as-you-go has no monthly fee. Paid plans include a monthly amount of routing; beyond that you continue at the flat lane rates. Exact plan amounts are on the pricing page.

No prepaid credit blocks to manage. Free includes $1 of routing per month. Paid plans include a monthly routing amount and then bill usage past it. Pay-as-you-go bills only what you route. Check the pricing page for the current plan details.

flux-auto is the default smart-routing model. Instead of pinning a model, you send "model": "flux-auto" and Flux picks a model sized to your prompt. It's the simplest way to use Flux and the one we recommend starting with.

Yes. Pass a specific alias as the "model" field instead of flux-auto. You can target a lane with flux-fast, flux-standard, flux-reasoning, or flux-image, or pin one exact backing model with its flux-pinned-* alias (37 are available). Pinned requests still go through the metered gateway.

Today routing is deterministic: each lane has a default model, and Flux maps your request to a lane based on the kind of work it is, then serves that lane's model. The goal is to send each prompt to the cheapest model that can do it well rather than defaulting everything to the most expensive option.

Yes. The model list grows as new models ship and prove out. Because you target stable aliases like flux-auto and the flux-* lane names, you generally pick up improvements without changing your code.

They're two wire formats for the same gateway. If your client speaks OpenAI, use POST /v1/chat/completions or POST /v1/responses at https://api.fluxrouter.ai. If your client speaks Anthropic, use POST /anthropic/v1/messages. Same key, same models, same routing underneath.

Yes, if they target the OpenAI or Anthropic APIs. Flux implements those shapes, so most tools work without code changes once you repoint them.

Usually just the base URL and the API key. Point the OpenAI client at https://api.fluxrouter.ai (or the Anthropic client at https://api.fluxrouter.ai/anthropic), use your sk-... key, and set the model to flux-auto or a flux-* alias.

Streaming and tool or function calling are supported on the standard OpenAI and Anthropic request shapes. Use them the same way you would calling the providers directly.

You get a single key in the format sk-... from your dashboard. It works across every endpoint and every model, passed as "Authorization: Bearer sk-..." (the Anthropic path accepts the same Flux key). Treat it like any provider secret: keep it server-side and out of client code.

Yes. Create and revoke keys from your dashboard. Rotate by issuing a new key, moving your apps over, and revoking the old one.

Your key is a bearer credential, so anyone who has it can spend on your account. Keep it in environment variables or a secrets manager, never in client-side code or a public repo, and revoke and rotate immediately if it leaks.

Flux is a routing gateway, not a model trainer. We do not build our own models from your prompts or responses. Your requests are forwarded to the underlying provider that serves them, and each provider has its own data-handling terms.

We keep the operational metadata needed to meter usage and bill you accurately. We don't sell your data, and we don't use your prompts or responses to train models. For specific data-handling questions, email hello@fluxrouter.ai.

Support is self-serve first. For live status and incidents, check https://status.fluxrouter.ai. For everything else, email hello@fluxrouter.ai. A Discord community is coming soon.

Every response includes X-Flux-* headers for transparency. x-flux-model tells you which model answered and x-flux-cost-usd reports the cost of that request, so you can verify routing and spend on a per-call basis.