OpenApps

FoxCut MCP server — tools & history

FoxCut MCP server: observed open status, tool schemas, linked apps and version history.

FoxCut

Endpoint: https://mcp.foxcut.ai/mcp

Observed status: open. Checked: 2026-10-13T05:06:44+00:00.

Linked apps

  • FoxCut

    AI video generator and image studio for Cursor: camera-directed video, text to video, image to video, talking avatars, consistent characters, product shots, and UGC ads. Kling, Sora, Veo, and the FoxCut catalog. Plan credits, then generate, edit, upscale, and assemble.

Observed tools

  • account_status — Use this when the user asks about their connected FoxCut account, plan, status, or credit balance. Do not use it to price one specific generation.
  • assemble_video — Use this when the user has confirmed the quoted credit cost and wants 1–20 completed clips assembled with cuts or crossfades and optional audio. Do not call it with unfinished job IDs.
  • cancel_generation — Cancel and refund only an unstarted queued job with no provider submission. Running, submitted, and completed jobs cannot be canceled or refunded by this tool. A delayed output is recovered rather than refunded. Requires an explicit user cancellation request.
  • create_brand_asset — Use this when the user has confirmed the quoted credit cost and wants one reviewable brand deliverable such as a logo concept, palette, typography, packaging, signage, or social identity.
  • create_character — Use this when the user wants to save a draft reusable identity from 1–8 private FoxCut source images. This step does not train the identity or spend training credits.
  • create_marketing_video — Use this when the user has confirmed the quoted credit cost and wants a UGC, how-to, unboxing, showcase, review, TV spot, wildcard, or virtual try-on video concept.
  • create_marketplace_cards — Use this when the user has confirmed the total quoted credit cost and wants 1–4 ecommerce card concepts from one product image. Each returned job is labeled by card type.
  • create_moodboard — Use this when the user wants to save a reusable visual style from private FoxCut reference images for later generations. It does not generate media.
  • create_video_explainer — Use this when the user has confirmed the total quoted credit cost and wants 1–8 explainer scenes from a structured plan. It returns asynchronous jobs that must be polled separately.
  • create_youtube_thumbnail — Use this when the user has confirmed the total quoted credit cost and wants 1–4 YouTube thumbnail or vertical-cover concepts with a mobile-readable composition.
  • edit_media — Use this when the user has confirmed the quoted credit cost and wants a natural-language change to existing image or video media while preserving the original.
  • estimate_generation_cost — Use this when you need the exact FoxCut credit charge before billed generation. It creates no job and spends no credits.
  • generate_audio — Generate standalone WAV speech after confirming the quoted FoxCut credit cost. Put only dialogue and sparse inline cues such as [calm] or [whispering] in transcript. Direction accepts one delivery preset, never acting paragraphs: Fish brackets are text and may be spoken. For dialogue use Name: lines with assigned voice IDs. speech_model selects Fish's model; omit to use the configured default. Success means audio was saved, not that spoken words or performance were verified.
  • generate_image — Use this when the user has confirmed the quoted credit cost and wants to create, edit, or restyle an image. This starts a billed asynchronous job; do not use it only to estimate cost.
  • generate_speech — Use this when the user has confirmed the quoted credit cost and wants a portrait to speak a script. This starts a billed asynchronous talking-avatar job.
  • generate_video — Use this when the user has confirmed the quoted credit cost and wants text-to-video, image-to-video, H3 MAX reference-to-video, or Kling motion transfer. Reference mode requires owned media, at most 12 files total, no start/end frames or character IDs. Name assets Image 1 / Video 1 / Audio 1 in the prompt. Quote kind=t2v with params.videoMode=reference and camelCase reference lists first. This starts a billed asynchronous job.
  • get_generation — Use this when you have a FoxCut job ID and need its current progress, terminal error, or final media URLs. Poll reasonably until the job finishes.
  • import_media — Use this when external provider-hosted image, video, or audio must become a private FoxCut asset before another tool can reference it. Do not use it for media already in the user's FoxCut library.
  • list_assets — Use this when the user asks for recent FoxCut image, video, or audio assets or another tool needs an existing asset URL.
  • list_characters — Use this when the user wants a consistent character and you need to find available trained or draft FoxCut identity IDs.
  • list_generations — Use this when the user asks for recent FoxCut jobs or you need to recover a job ID. Use get_generation when one known job needs detailed polling.
  • list_models — Use this when choosing or comparing a FoxCut image, video, or talking-avatar model. It returns capabilities, limits, and credit rates; it does not create media.
  • list_presets — Use this when the user wants a camera move, visual style, or marketing template and needs valid FoxCut preset IDs before generation.
  • list_voices — Find real voice IDs and playable samples for speech generation. Includes your private clones. Defaults to licensed voices; community voices are optional.
  • list_workflows — Use this when the user wants a complete creative outcome and you need to discover FoxCut's multi-step workflows before choosing individual generation tools.
  • prepare_subtitles — Transcribe an owned existing video locally or use supplied timed cues, compare authored dialogue, and pair supplied translations. Returns a draft job ID; poll get_generation for cues and review warnings. No video generation or automatic translation is called. Requires an enabled media worker.
  • render_subtitles — Render one or two subtitle languages together from the original video and a completed draft. Use preview=true to validate all glyphs/layouts and return representative cue frames. Review warnings before export; audio transcoding requires allow_audio_transcode. Returns a job ID; get_generation returns video, SRT/VTT per language, cue JSON and technical QA. Optional official-logo end card is composed deterministically.
  • run_product_photoshoot — Use this when the user has confirmed the quoted credit cost and wants a commercial product-photo concept from a source product image and a purpose-built shot mode.
  • train_character — Use this only after the user explicitly confirms the 350-credit charge for training a draft FoxCut character. Check account_status first.
  • upload_media — Read a local file and send its base64-encoded bytes to create a private FoxCut asset. Maximum 5 MiB. Use the returned asset URL as a generation input. For files up to 50 MB use POST /api/mcp/upload with the same OAuth bearer token and multipart file/purpose fields. File paths themselves cannot be read by this remote server.
  • upscale_media — Use this when the user has confirmed the quoted credit cost and wants a higher-resolution copy of existing image or video media. It does not overwrite the source.

Data as of 2026-10-13T05:06:44+00:00.

OpenApps records store listings and observed changes. Store ranks are not download counts; launch dates may be derived or use the first observation. Coverage gaps are shown in the interactive catalog.

REST API specification · Guide for AI clients · Manage API keys and MCP connections