Fire is the inference engine underneath Modus Promethean: a single, model-agnostic API that routes chat, images, speech, transcription, decisions and multi-agent flows to the right provider — and returns the exact price of every call before you've finished reading the response.
Most "AI gateways" are a thin proxy in front of one provider's SDK. Fire is the pipeline every app in the Modus Promethean stack — Saga, FireScout, ImageLab, PornArt — runs production traffic through. What's sold here is exactly what's dogfooded.
One request shape for 53+ models across Anthropic, OpenAI, Google, Mistral, DeepSeek, xAI and
more. Swap species_name, not your integration.
Every response carries meta.price.usd — the amount you were actually charged, even when it's
zero. No reconciliation, no surprise invoice.
Compose DAGs of model calls with parallel branches and real human gates. Queued, resumable, billed per step — and designed visually in the portal.
Flows →Metadata is always logged — model, tokens, price. Prompt and response content is stored only if you opt in, with a retention window you choose.
Data policy →Fire doesn't pick a lane. Each endpoint routes to the model best suited for the job and is metered the same way regardless of what's on the other end of the wire.
Frontier and fast-tier text models, streaming, parallel fan-out, tool calls and reasoning budgets handled for you.
Flat per-image pricing, reference-image edits, upscaling, and a compliance-checked Layer 2 workflow when you want the whole pipeline.
Text-to-speech across five voices engines and Whisper-class transcription — the two halves of a voice agent.
Typed answers with probabilities — yes/no, choice, score — for the places a free-text model shouldn't be trusted to be consistent.
Named agents, dependency graphs, repeat macros, web search budgets and human gates, run asynchronously and billed per step.
Need a third-party service? Fire injects the credential server-side and forwards the call. Your token is the badge; it never sees the key.
Get a key from the portal, then you're one request from a response — with the price you were actually charged and whether the content was logged returned inline, every time.
Discovery is public: GET /v1/capabilities describes every endpoint, model and policy with no auth, so an agent can bootstrap itself.
A real call succeeds right now, usage is reported, cost is verified, a price is published. A weekly audit re-runs the same checklist — a model that dies upstream is pulled automatically, not left silently broken.
Buy a credit pack, spend it at your own pace. Text is billed off a published per-model rate, images are a flat per-image price, and every call returns the amount it cost.
Free to sign up — you only pay when you top up. No card required to get your first key.