MCP
MCP tool reference
Current FoxCut MCP tools, their billing behavior, and result contract.
Read-only tools
| Tool | Purpose |
|---|---|
| account_status | Plan, status, identity, and credit balance |
| list_models | Model capabilities and rates |
| list_workflows | Task-oriented workflow catalog |
| list_presets | Motion, style, and marketing presets |
| list_voices | Standalone speech and talking-avatar voice choices |
| estimate_generation_cost | Credit estimate without submission |
| get_generation | One job’s status and outputs |
| list_generations | Recent jobs |
| list_assets | Generated media |
| list_characters | Reusable identities |
H3 MAX reference-to-video
Use generate_video with model=minimax-h3-max, video_mode=reference, and reference_image_urls, reference_video_urls, or reference_audio_urls. Import or upload media first and pass the returned relative asset content URLs. References guide identity, style, movement, or sound; they are not first/last frames.
- Up to 9 images, 3 video clips, 3 audio clips; at most 12 files total. Video/audio clips must each be 2–15 seconds, with a 15-second total for each modality. Audio requires at least one image or video.
- Name assets in list order: Image 1, Image 2, Video 1, Audio 1. Describe each role, then the action, cuts, and which character says each line. Following these directions is not guaranteed.
- Do not mix reference mode with start/end frames, character IDs, style/camera presets, or the singular reference_video_url (Kling motion transfer).
- Estimate kind=t2v with params.videoMode=reference, referenceImageUrls, referenceVideoUrls, referenceAudioUrls, durationSec, resolution, and aspect. The estimate includes reference costs and reads verified asset metadata; client token counts are ignored.
- Use 480p or 768p, the durations/aspects in list_models, and standard or draft quality. Generation reserves the quoted credits and returns a job ID to poll.
Creation tools
| Tool | Behavior |
|---|---|
| generate_image | Queues text-to-image or image edit |
| generate_video | Queues text/image/reference/motion-to-video |
| generate_audio | Queues standalone WAV speech; 2–8 credits by emitted UTF-8 bytes |
| generate_speech | Queues talking-avatar video |
| edit_media / upscale_media | Non-destructive derivative jobs |
| run_product_photoshoot | Commercial image modes |
| create_marketing_video | UGC, demo, review and ad modes |
| create_brand_asset | One selected brand-system deliverable |
| create_marketplace_cards | One or more commerce cards |
| create_video_explainer | Narrated scene or assembled explainer |
| create_youtube_thumbnail | 16:9 or vertical cover image |
| create_character / train_character | Draft and train identity |
| create_moodboard | Save reusable visual direction |
| cancel_generation | Stops eligible work; does not delete assets |