Stability AI

Stability AI

This connector uses the Stability AI REST API (https://api.stability.ai) to create and edit pictures with the Stable Image services (Ultra, Core, Stable Diffusion 3.5, upscale, edit, control) and to create music with Stable Audio 3. Finished pictures are attached to the tool result as images the user can see; audio is attached as a file.

Setup (3 minutes)

1. Create an account at https://platform.stability.ai (new accounts get 25 free credits) and buy credits under Billing if needed.
2. Open API keys (https://platform.stability.ai/dashboard/api-keys), copy a key (it starts with sk-) and paste it into STABILITY_API_KEY. It is sent as Authorization: Bearer.
3. The key acts for the default organization of its user; credits are taken from that organization.

Costs: every generation, edit, upscale and audio tool COSTS CREDITS (1 credit = $0.01): Core 3, Ultra 8, SD 3.5 Large 6.5, Large Turbo 4, Medium 3.5, Flash 2.5, fast upscale 2, conservative upscale 40, creative upscale 60, erase, inpaint, search and replace, search and recolor, remove background 5 each, outpaint 4, replace background and relight 8, sketch, structure and style 5 each, style transfer 8, Stable Audio 3 8. Failed generations are free. Tell the user the cost and confirm before every paid call; check stability_ai_get_balance first when in doubt. Reads (stability_ai_get_balance, stability_ai_get_account, stability_ai_list_engines, stability_ai_get_result, stability_ai_get_audio_result) are free.

Which tool

• Create from text: stability_ai_generate_image_core (fast, cheapest, good default), stability_ai_generate_image_ultra (best quality, photorealism, also from a starting picture with strength), stability_ai_generate_image_sd3 (Stable Diffusion 3.5 models; mode image-to-image with image_url and strength).
• Improve resolution: stability_ai_upscale_fast (4x, cheap), stability_ai_upscale_conservative (to about 4 megapixels, faithful), stability_ai_upscale_creative (adds detail, asynchronous).
• Edit: stability_ai_remove_background, stability_ai_erase (remove objects with a mask), stability_ai_inpaint (repaint a masked area), stability_ai_outpaint (extend the canvas), stability_ai_search_and_replace and stability_ai_search_and_recolor (find an object by words, no mask), stability_ai_replace_background_and_relight (asynchronous).
• Guide with a picture: stability_ai_control_sketch (drawing to picture), stability_ai_control_structure (keep composition), stability_ai_control_style (copy a style), stability_ai_control_style_transfer (style of one picture onto another).
• Audio: stability_ai_generate_audio (music and sound effects, asynchronous), then stability_ai_get_audio_result.
• Account: stability_ai_get_balance, stability_ai_get_account, stability_ai_list_engines (legacy v1 engines only).

Input pictures: every *_url parameter is a public https link to a JPEG, PNG or WebP file (up to 10 MB). AnythingMCP downloads it and uploads it to Stability AI as a file; a picture the user only pasted into the chat has to be put online first. Masks are pictures where white marks the area to change; without a mask the picture's transparent areas are used.

Asynchronous jobs: stability_ai_upscale_creative, stability_ai_replace_background_and_relight and stability_ai_generate_audio return { "id": "..." } (64 characters) at once. Call stability_ai_get_result (pictures) or stability_ai_get_audio_result (audio) with that id after about 10 seconds; status: in-progress means call again later. Results stay available for 24 hours and only for the same key.

Reading results: picture tools return finish_reason and seed with the attached image. finish_reason CONTENT_FILTERED means the output broke the content policy and was blurred; rephrase the prompt. Reuse seed with the same prompt to reproduce a picture. Pictures default to JPEG to stay small; ask for png or webp when transparency is needed.

Prompts: write in English, describe subject, setting, style, colours and lighting; put what to avoid in negative_prompt. style_preset (photographic, cinematic, anime, digital-art, 3d-model ...) nudges the look.

Errors: 400 lists the wrong parameters in errors; 401 bad key; 402 not enough credits; 403 the prompt or picture was flagged by content moderation; 413 the upload is too large; 422 the input cannot be processed (e.g. picture size out of range or unsupported language); 429 rate limit (150 requests per 10 seconds, then a 60-second block). Every error carries an id to quote to Stability support.

Out of scope: 3D model generation (Stable Fast 3D, SPAR3D), audio-to-audio and audio inpainting, the Stable Audio 2 endpoints and legacy v1 SDXL generation are not part of this connector.