Real output
The output is a URL you keep.
Not a blob that vanishes with the shell. You get a hosted image URL, a job you can look up later, and a receipt for what it cost. These were generated through the public CLI, no key involved.
COVER
MACRONo provider key. No OAuth screen. No billing console to set up before your first image. Luxin's CLI signs your agent up on the first run, picks a model for the prompt, and returns an image URL your agent owns. A free preview covers the first images; you add prepaid credits only when you want more.
Full disclosure: Luxin is our product, and this is our own guide. We wrote the honest version, including where it is the wrong tool.
The usual first hour of image work is registering with a provider, copying a key into an env file, and turning on billing. This path skips all of it.
The first run is a no-spend guide pass. It checks that Luxin is reachable, picks a runnable model for your prompt, and prints the exact create command. Nothing is charged.
There is no key to paste and no OAuth window to click through. The CLI creates the account on the first run and stores its own token, so the next command just works.
A free preview quota covers your first images. Past that, credits are prepaid at 1 credit for 1 cent, and every job prints a quote before it spends. You set a cap, so nothing runs an open-ended bill.
Not a blob that vanishes with the shell. You get a hosted image URL, a job you can look up later, and a receipt for what it cost. These were generated through the public CLI, no key involved.
COVERMACROA script or an agent needs an image and you do not want to stop to register for a provider, copy a key into an env file, and turn on billing. You want the result to be durable work with a URL, a job, and a receipt, and you want a spend cap and a clear bill.
If a provider SDK and key are already set up and working for you, a second account buys little. And if you need generation that runs fully offline or on your own hardware today, Luxin is hosted, so it will not fit that yet.
Luxin holds the provider relationships and keys on its side. Your agent authenticates to Luxin, not to each provider, and the CLI sets that up on the first run. You never register with a model provider or paste a key.
There is no signup form to fill in. The CLI creates the account on the first run and stores its own token, so nothing blocks your first image. A human adds credits later, only once you want more than the free preview.
There is a free preview quota so you can try it before paying anything. After that, credits are prepaid at 1 credit for 1 cent, priced per model, and every job shows a quote before it spends. There is no subscription and no charge you did not approve.
You fund the Luxin account, not a provider. A human can top up with one Stripe click, or the agent can pay for its own runs within a cap you set. Either way it is prepaid, so a run cannot go past the balance.
It removes one. There is no provider secret sitting in your shell history or an env file to leak. The agent's Luxin token is scoped to its own account and spend cap, and every job and receipt is visible to you.
Luxin auto-selects a model that fits the request from more than 200 across providers, and can write the prompt for you. If you want a specific model, you can name it; the default is there so you never have to.
No. The same command and contract cover image generation, image editing, video, audio, and image-to-3D, and they run the same from a shell script, CI, or any agent you point at luxin.sh/skill.md.
Want your agent to always have it? Install the skill: npx skills add https://github.com/danielgwilson/luxin --skill luxin