Generate images on behalf of an app's own users
No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.
What counts as done
Unsafe requests refused with a machine-readable reason before billing, and spend per user capped.
Measured, for our own tests: safe requests return within the app's latency budget. No docs we read state these figures, so they don't count against a listing here.
How agents ask for it: let my users make avatars
, add AI images to my story app
, let visitors redecorate the room
Does the job, with a limit 9
It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.
| Listing | Why |
|---|---|
| Black Forest Labs FLUX API B Black Forest Labs Inc. | Request Moderated fires before processing with a Moderation Reasons array, and safety_tolerance 0 to 7 sets strictness, but whether moderated requests are billed isn't stated and no per-user spend cap exists. source |
| Bria API BB Bria Artificial Intelligence Ltd. | prompt_content_moderation and content_moderation are opt-in and refusals return 422 'Prompt failed content moderation', but no category code, no statement on billing a refusal and no per-user spend cap. source |
| fal image models BB Features & Labels, Inc. | Models take safety_tolerance and flagged inputs return a non-retryable 422 content_policy_violation, but with no reason category, 422s may still be charged, and no per-user spend cap was found. source |
| FLORA B FLORA AND FAUNA FLOWERSHOP, INC. | Moderated prompts fail the run with GENERATION_PROMPT_MODERATED and admins can set member spend limits, but the refusal arrives after the run starts with no category, and limits apply to workspace seats, not an app's end users. source |
| Ideogram API D Ideogram AI, Inc. | Unsafe prompts return 422 and outputs carry is_image_safe, but there is no category code or moderation level, dry_run skips the safety review, and no per-user spend cap. source |
| Leonardo.Ai API F Leonardo Interactive Pty Ltd | NSFW prompts are blocked by default with a 400 naming the matched term, and outputs carry an nsfw flag, but the code field is 'unexpected', stricter controls need contact with sales, and no per-user spend cap exists. source |
| OpenAI Image API BB OpenAI OpCo, LLC | moderation auto or low, and moderation_blocked with stage and categories, give the app a machine-readable refusal, but no page says whether a blocked call is billed and no per-user spend cap was found in the pages read. source |
| Seedream on BytePlus ModelArk C BytePlus Pte. Ltd. | A moderated image in a sequential batch returns its own error and isn't billed, but the error carries a message for troubleshooting rather than a category, there is no moderation level and no per-user cap. source |
| Stability AI Image API F Stability AI Ltd | A moderated request returns 403 content_moderation and states it wasn't charged, and outputs can come back CONTENT_FILTERED, but there is no category or moderation level and no per-user spend cap. source |
Couldn't check 2
The docs were closed to us, or silent. These carry no mark either way.
| Listing | Why |
|---|---|
| Adobe Firefly API D Adobe Inc. | The OpenAPI file names no moderation error or reason code, and the usage notes don't describe refusals; nothing to judge from. |
| Qwen-Image 2.1 (Model Studio) B Alibaba Cloud (Singapore) Private Limited | Moderation and error codes live on Model Studio pages that didn't resolve from here. |
Doesn't do this job 4
Listings in the category that do something else, with the reason.
| Listing | Why |
|---|---|
| Iris-3B C Speridlabs | No moderation, refusal codes or spend controls are documented; the card says outputs can be inaccurate or unsafe and must be reviewed. source |
| Recraft API D Recraft Inc. | The error page lists only 401 and 404, with no moderation reason codes, moderation setting or per-user spend cap. source |
| Replicate image models B Replicate, LLC | Safety is left to each model; the platform documents no refusal reason code, and tokens carry no spend cap, so per-user caps must be built outside. source |
| SurgePix API F SurgePix | Uploads pass NSFW detection, but no refusal codes, moderation settings or per-user spend controls are documented. source |
What stops agents finishing
- no input moderation so abusive uploads reach the model
- refusals give no reason code the app can show
- moderation too strict for legitimate fashion or art
Around it
Before: auth.user-session. After: storage.upload.
Where this comes from
Weight 6 of 100 in Image generation: 3 of 37 accounts; an image app overrun by NSFW uploads, a public room-edit toy degrading on fal, a story app on Replicate
The same as data: /api/v1/use-cases/image-generation.end-user-generation.json, or the MCP tool get_use_case with id image-generation.end-user-generation. Other jobs in this category: /api/v1/niches?category=image-generation.
For companies
Do agents find, use and choose your tools?
An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.