Workflows that aren’t only text
Plenty of real team workflows end in something visual or audible — a blog hero, a social card, a generated jingle. Forge treats media as a first-class output: agents call one generic generations API and get back a stored artifact, the same way they call any other tool.
How it works
The generations API is kind-parameterized — you request an image (with audio and more on the same surface) and a prompt. You don’t pick a model; the multi-provider gateway
selects a capable provider from the existing pool, using the same credentials, routing, and health tracking it already uses for chat. No new provider integration, no separate key to manage.
The generated binary is written to Forge object storage and returned as a self-authenticating link, so a large image or audio file never has to be inlined into a transcript or a handoff.
We dogfood it
Every hero image on this blog was generated through this exact API — one consistent auth and storage layer, provider chosen by the gateway. If it’s good enough for our own marketing, it’s wired into your workflows too.