Higgsfield MCP

Read-only catalog of the 3D rig animation library (678 actions: locomotion, gestures, dancing, combat, daily actions). Search by name or browse by group/category to find the animation_action_id for 3D generation with enable_animation=true. Each result has a preview_url GIF — when several candidates fit (e.g. many Idle or Walk variants), show the user the previews as markdown images and let them pick instead of choosing blindly. Does not create jobs.

Get an app's action contract: with `action`, the full input/output schema + execution mode for that one action; without it, a summary of every action. Also returns `manifest_revision`, which apps_invoke requires. Read-only.

Run one described action on a Marketplace app AS the current user. First call apps_describe(app_id, action) to get the exact `arguments` schema and the `manifest_revision`, then pass them here. Long-running actions return { id, status: "queued" }. If a widget is visible, it polls the status action automatically — do not re-invoke get_* as a follow-up poll. In text-only clients, poll by invoking the app's status action (e.g. get_render) until status is completed/failed. The action's own annotations (from apps_describe) indicate cost/side-effects; confirm with the user before an expensive or destructive action.

Search Higgsfield Marketplace apps callable through MCP. Returns each app's id, name, and the actions it exposes. Flow: apps_search to find an app → apps_describe(app_id, action) to get an action's argument schema + manifest_revision → apps_invoke to run it. Read-only; does not call any app.

Get the user's available credits and current subscription plan. For transaction history, call `transactions` instead.

Cancel the auto-renewal of the Higgsfield MCP free trial. Call this when the user asks to cancel the trial, cancel auto-renewal, stop the upcoming charge, or asks how to cancel. IMPORTANT SEMANTICS: cancelling stops the automatic charge at the end of the trial ONLY — the user KEEPS trial access and remaining trial credits until the trial ends; nothing is charged. First call WITHOUT `confirm` (or confirm=false): in UI clients this opens a confirmation card with 'Keep trial with auto-renewal' and 'Cancel auto-renewal' buttons; in text-only clients relay `assistant_response` verbatim and wait for the user's explicit confirmation. Only call again with `confirm=true` after the user explicitly confirmed the cancellation in chat. Never pass confirm=true on the first call.

INTERNAL — invoked ONLY by the plans widget on an explicit user Confirm click. Do NOT call this tool yourself; it charges the user's real saved payment method off-session. To help a user upgrade, top up, or set auto top-up, call `show_plans_and_credits` instead — it returns checkout links and, in UI clients, opens the widget where the user confirms the charge. Covers auto_topup and topup via the `action` field (upgrade_plan is temporarily disabled — plan upgrades use the hosted checkout link). The backend either charges the saved card (`result: "charged"`) or returns a `checkout_url` to redirect (`result: "redirect"`). Buying a brand-new subscription is not handled here — that uses the hosted checkout link.

INTERNAL — invoked ONLY by the cancel-trial confirmation widget on an explicit user click of 'Cancel auto-renewal'. Do NOT call this tool yourself; to cancel the trial's auto-renewal use `cancel_trial_auto_renewal` instead. Stops the trial's auto-renewal (cancel_at_period_end) — trial access and remaining credits stay until the trial ends.

Open the Create Voice Apps UI. Call this immediately when the user asks to create a voice, call the Create Voice tool, or needs a local browser record/upload surface and no confirmed audio_media_id is already present. Do not ask the user to upload an audio file or provide the name in chat first; the widget collects the required name plus record/upload audio. If the user already has or attached an audio file in chat, still call this tool with initial_tab='upload' — remote tools cannot read Claude chat attachments, so the user re-selects the file in the widget's Upload tab (it uploads directly to Higgsfield). Do not try to pass a chat attachment or ask for a URL. The widget records/uploads, confirms the audio, and creates the voice itself end-to-end — and if the user is out of credits or on a free plan it shows the plans/credits UI inline. After the widget reports success you do NOT need to call create_voice_from_confirmed_audio again. Only call create_voice_from_confirmed_audio yourself when a confirmed audio_media_id is already present in the prompt and no UI step is needed.

Backend-only creation of a cloned voice from an already confirmed audio upload. Do not call this tool until audio_media_id and name are already known. For direct creation, first upload speech audio with media_upload, PUT the bytes, then call media_confirm with type='audio'. Pass that confirmed media_id here as audio_media_id plus a required name. If the user needs to record or upload local audio in an Apps UI-capable client, call create_voice instead; that widget records/uploads, confirms, and creates the voice itself, so you do not call this tool for the UI flow. The audio should be clear speech, roughly 10 seconds to 3 minutes, and no larger than the upload limit. The backend charges the voice-clone credit cost on successful creation. Cloning is asynchronous: on success the tool returns the new voice_id plus a status, and a fresh clone is usually still 'processing' and not yet usable. Use the returned voice_id with voice_type='element' for generate_audio or voice_change only once it is ready (status='completed' and is_audio_eligible=true). If status is 'processing' or is_audio_eligible is not true, the clone is still training — re-check it with list_voices before generating instead of submitting right away; status 'voice_clone_failed'/'failed' means cloning did not succeed. If recovery_tool is returned, call it immediately; do not explain/ask first.

Start a new full-stack website. Creates the website and a git repo: a React 19 + TanStack Start app, server-rendered, in ONE Cloudflare Worker, with D1 / R2 / KV / Durable Objects / Containers available (all DISABLED by default). Returns a website_id — pass it to every later website tool. The 'type' param is REQUIRED and is the USER'S choice, not yours: unless the user has already made it unambiguous, ASK the user whether they want a plain website (no Higgsfield integration) or a Higgsfield-integrated app (Sign in with Higgsfield + AI image/video generation via the Higgsfield SDK) BEFORE calling this tool. Apps are scaffolded from a v2 starter template and REQUIRE the 'template' param — pick the closest of studio / preset / app-detail per the template param's guide ('custom' is ONLY for when the user explicitly says "use custom template" — never pick it yourself). The chosen layout ships as real code already wired as the home page; you ADAPT IT IN PLACE, never rebuild it. Websites take an OPTIONAL template: pass 'scroll-scrub' for an animated website (its scrub engine ships pre-built) and omit it for a non-animated one. App and website templates are not interchangeable — a cross-kind name is rejected. Workflow: (0) call get_workflow_instructions with { workflow: "website-builder-flow" } FIRST to load the stack, design contract, and hard rules (REQUIRED before building or editing); (1) create_website; (2) call website_repo_access to get the repo's git URL + scoped token, clone/edit/commit/push with the terminal (for apps, read app/src/layouts/AGENTS.md + app/src/components/AGENTS.md right after cloning); (3) deploy_website to ship it live — and deploy again after ANY later change (publish_website only lists what is already live; it does not deploy).

Build and deploy the website via CI, then return its live URL. Every deploy ships the live site at the website's public URL (there is no separate preview stage). IMPORTANT: commit and git push ALL your changes BEFORE calling this — the build runs from the pushed repo. Deploy again after ANY later change: publish_website does NOT deploy (it only lists the already-live build on the community feed), so this tool is the only way changes ship. A failed build returns the log; a still-running build returns status 'pending' — call website_status to check.

Dub a video into another language: translate the spoken audio, synthesize it in the target language, and lip-sync the result back onto the video. Use this when the user asks to dub, translate the speech of, or localize a clip into another language. Pass video_id for the source video (a confirmed uploaded media_id or a completed video generation job_id) and target_language as one of the supported language codes. Supported languages (code=language): eng=English, cmn=Chinese, fra=French, hin=Hindi, ita=Italian, jpn=Japanese, kor=Korean, por=Portuguese, rus=Russian, tur=Turkish, spa=Spanish, deu=German, ara=Arabic, pol=Polish, ind=Indonesian, fil=Filipino, swe=Swedish, fin=Finnish. This tool does not use prompt or count; output dimensions are taken from the source video automatically.

Generate a 3D GLB mesh. Use `models_explore(type:'3d')` to pick a model and see its `medias[].roles` and `parameters`. Apps UI local file: call `media_upload_widget`; remote tools cannot read Claude chat attachments. Web media URL: call `media_import_url`, pass returned `media_id`; `medias[].value` must be media_id/job_id, not URL. Defaults: `image_to_3d` for general image-to-3D with optional texturing, PBR, and rigging; `multi_image_to_3d` when 2-4 views of the same subject are available (better geometric accuracy); `sam_3_3d` for single-object reconstruction; `3d_rigging` to rig an existing 3D model (takes `model_url`, not images — pass a prior 3D job_id or an https GLB URL). For animated rigs, search clip ids with the `animation_actions` tool and pass `animation_action_id` with `enable_animation:true`. The mesh reproduces only what is in the source image — to add or change props, clothing, or held objects, edit the image first with `generate_image`, then convert the edited result. Pass model-specific params as top-level fields. Apply `adjustments` returned by the server. If `recovery_tool` is returned, call it immediately. `get_cost:true` preflights credits without submitting.

Generate one speech/voice request (text-to-speech) and render it in the generation widget. This tool accepts one prompt; for 2-12 independent lines or prompts, use the headless generate_audio_batch tool instead. DEFAULT model: seed_audio (Seed Audio 1.0 by ByteDance) — use it unless the user explicitly asks for a different engine. seed_audio takes a preset or reference-element voice (voice_type 'preset'|'element' + voice_id) plus optional tuning params (format, sample_rate, speech_rate, loudness_rate, pitch_rate), and can clone a voice from an audio_references media item or take an image_references cue. To use a specific named engine instead, set model:'text2speech_v2' and pass variant (one of elevenlabs|minimax|seed_speech|vibe_voice|cozy_voice) together with voice_type + voice_id. Get voice ids from list_voices; use models_explore(type:'audio') to inspect each model's params. This tool only generates speech: it cannot generate music or sound effects for general use, and there is no standalone music/SFX model here — decline general music or sound-effect requests rather than substituting a speech model. The models sonilo_music (music), mirelo_text_to_audio (sound effects) and inworld_text_to_speech (voice) exist ONLY for the game-generation pipeline and must not be used for standalone audio. get_cost:true preflights credits without submitting. use_unlim defaults false — pass true only when the user explicitly asks to use their unlimited/free-trial generations, never to save them credits on your own initiative.

Submit 1-12 independent audio generations in parallel without opening a widget. Each requests[] item accepts the same params as generate_audio, creates exactly one job, and keeps its caller-provided index in the response. Use for multiple distinct prompts or inputs; use generate_audio for one user-facing generation. Poll returned job IDs with jobs_wait in agent-chosen groups of at most 12. For larger sets, collect indexed jobs across submission batches. After every job in the user's set is terminal, pass the collected jobs to exactly one show_generation_by_ids call for up to 60 jobs; never use show_generations or call job_display once per job.

Generate one image request and render its result(s) in the generation widget. Use count 2-4 only for variants of the same prompt, inputs, and settings; for 2-12 independent image requests with different prompts or inputs, use the headless generate_image_batch tool instead. Apps UI local file media: call `media_upload_widget`; do not ask for Claude chat attachments because remote tools cannot read them. Web media URL: call `media_import_url`, then pass returned `media_id`; `medias[].value` must be media_id/job_id, not URL. Default general image model: `gpt_image_2` — use it for ordinary generation, photorealistic images, typography, and reference-based editing unless a specialized route applies. Specialized defaults: `marketing_studio_image` for commercial/product/ads; `soul_cast` for text-only character/avatar; `soul_2`+`soul_id` for trained reusable Soul; `soul_2` for portraits/fashion/UGC/editorial. Ambiguous create-character/avatar: offer reusable Soul training (5-20 photos, ~10 min) vs one-off; do not train generic silently. Use `show_characters(action='train')` only if explicitly requested or user provides 5-20 photos. Use `models_explore` for aspect_ratios, params, medias roles. Top-level model params; apply `adjustments`. If `recovery_tool` returned, call it immediately; do not explain/ask first. `get_cost:true` preflights credits. `use_unlim` defaults false — pass true only when the user explicitly asks to use their unlimited/free-trial generations, never to save them credits on your own initiative.

Submit 1-12 independent image generations in parallel without opening a widget. Each requests[] item accepts the same params as generate_image, creates exactly one job, and keeps its caller-provided index in the response. Use for multiple distinct prompts or inputs; use generate_image for one user-facing generation. Poll returned job IDs with jobs_wait in agent-chosen groups of at most 12. For larger sets, collect indexed jobs across submission batches. After every job in the user's set is terminal, pass the collected jobs to exactly one show_generation_by_ids call for up to 60 jobs; never use show_generations or call job_display once per job.

Generate one direct video request and render it in the generation widget. Use count 2-4 only for variants of the same prompt, inputs, and settings; use headless generate_video_batch for 2-12 independent requests. GENJUTSU TRIGGERS: route `Higgsfield Genjutsu` by intent. Copy, repeat, reproduce, mimic, or transfer motion, movement, actions, gestures, dance, or camera motion from one driving video to reference-image subjects -> `hf_mult_motion_control`. Replace, change, or swap an object, product, garment, or character in one source video from reference images -> `hf_mult_replace_object`. These are direct `generate_video` models, not legacy `motion_control` or `ad-multiplier`; reserve ad-multiplier for explicitly requested independent variants. Pass images with role `image` and exactly one source/driving video with role `video`. LOCAL/ATTACHED INPUT GATE: without confirmed media_id values, call `media_upload_widget` first as the only tool in that turn; never inspect /mnt/user-data/uploads, run shell, or ask for a chat attachment. For mixed image+video use type:`auto`, multiple:true. For web media call `media_import_url`; `medias[].value` must be media_id/job_id. Defaults: `marketing_studio_video` for ads/products, `clipify` for YouTube clips, `seedance_2_5` for general video, `kling3_0` for multi-shot, audio, or motion transfer, and `minimax_h3` for 2K keyframes or mixed references. Marketing Studio: fetch URL products with `show_marketing_studio(action='fetch')`; create uploaded-image products with type `product`. List missing hooks/settings before presets. Use declared media roles and model-supported audio only. Use models_explore for durations/params. Apply adjustments and immediately call any recovery_tool. get_cost:true preflights credits. Set use_unlim:true only when explicitly requested.

Submit 1-12 independent video generations in parallel without opening a widget. Each requests[] item accepts the same params as generate_video, creates exactly one job, and keeps its caller-provided index in the response. Use for multiple distinct prompts or inputs; use generate_video for one user-facing generation. Poll returned job IDs with jobs_wait in agent-chosen groups of at most 12. For larger sets, collect indexed jobs across submission batches. After every job in the user's set is terminal, pass the collected jobs to exactly one show_generation_by_ids call for up to 60 jobs; never use show_generations or call job_display once per job.

Show the explainer video style presets (CMS-managed catalog). Returns preset ids, names, and preview media. When the user picks one, resolve it with resolve_explainer_preset to get the style reference media_id for generations.

Read a safe text file or directory from a workflow's resource folder. Use this after get_workflow_instructions when the SKILL.md requires a template, reference, or script file.

Ad Multiplier — load workflow 'ad-multiplier' when the user asks to 'multiply my video', 'multiply my ad', create multiple independently edited versions of one supplied 4-30 second video, or regenerate the same ad with different people or products. Load this workflow before Marketing Studio, model browsing, or direct generation. Brand Asset Creation: for branded-asset work including logo recoloring/export, a branded PowerPoint/presentation deck, or analyzing an official brandbook to produce an asset, even when all inputs are supplied or no generation is needed, load 'brand-asset-creation' before sandbox_exec. Faceless video generation, AI-narrated video, narrated animated explainer video, narrated / personal / philosophical story video, YouTube/Instagram thumbnail or video cover, product photoshoot, packshot, studio or lifestyle product photography, product hero banner, product carousel, static product ad pack, virtual model product try-on, conceptual product still, or product-photo restyle, UGC-style ad for a website / SaaS / store / product page from its URL ('SaaS UGC'), any other UGC / creator-style short video for a product — a talking-head creator review (the default UGC ask), a product-only ad with no creator on camera, an unboxing / first-reaction / haul, a try-on / fit check / OOTD, a step-by-step tutorial with on-screen steps, a character sheet, character reference, model sheet, turnaround, expression sheet, or consistent multi-view character prompt, any branding work — a logo, visual identity, brand kit, brandbook, branded mockups, merchandise, packaging, signage, social graphics, posters, or banners ('brand-asset-creation'), including recoloring or exporting an existing official SVG/PNG logo even when no new design or image generation is requested, or building / editing a website, web app, landing page, or browser game with the website tools ('website-builder-flow'): before building ANY of these, use this tool to discover and load the bundled workflow (each a SKILL.md that orchestrates the generate_* tools). Call with NO argument to list available workflows and their triggers. Call with a workflow name to load that workflow's full SKILL.md plus the list of files readable via get_workflow_bundle_file.

Show one specific previous generation in the single-result UI widget by job ID. Use when the user wants to inspect or re-display that individual result, including workflows that require separate approval of named candidates or individual previews before finalization. Do not call job_display once per job merely to reproduce an ordinary completed batch; use one show_generation_by_ids call for ordinary batch results instead.

Check the status and results of an async job. Returns instantly. For non-terminal jobs the response includes poll_after_seconds — wait that many seconds before calling again. Typical total times: image ~10-20s, video ~60-180s.

Long-poll 1-12 generation jobs together without opening a widget. Waits up to timeout_seconds (default 15, max 15) for every job to reach a terminal state, then returns compact indexed statuses and result URLs. Use job IDs returned by generate_image_batch, generate_video_batch, or generate_audio_batch. For larger sets, choose groups of at most 12 and wait for each group. Permanent lookup failures are returned once without blocking the other jobs; transient lookup failures are retried within the timeout. When all_terminal is false, wait poll_after_seconds before calling again. After every wait group in the user's generation set is terminal, collect their indexed jobs and display them with one show_generation_by_ids call when within that tool's limit. Never use show_generations or call job_display once per batch job.

List available voices for speech and voice tools. Returns built-in preset voices plus the user's own custom voices. Each voice has a voice_id and a voice_type ('preset' or 'element'); pass that exact pair to the audio models (via generate_audio — seed_audio or text2speech_v2) and to the voice_change tool to select the speaking voice. Use the preview_url to hear a sample. Paginate with the returned next_cursor.

List the content categories a website can be filed under — each with a slug, label, description, and display position. create_website REQUIRES a `category`; call this first to get the valid slugs, then pass the closest one ('other' when nothing fits).

List the websites you own — each with its id, name, slug, and live URL. Use this to find the id of a website you created earlier so you can edit, deploy, or check its status.

List every workspace the user can access (their private workspace plus any shared/team workspaces). The `is_selected` field marks which workspace MCP operations currently target. Use when the user asks which workspaces they have, or wants to switch workspace.

Widget-internal: list the user's Marketing Studio avatars (preset and custom) for the avatar picker.

Widget-internal: the Marketing Studio v2 pricing document. credits = cost_units / cost_units_per_credit; video flows cost fixed_cost_units + cost_units_per_second × duration.

Widget-internal: recreate a Marketing Studio v2 preset — validates inputs against the preset's recreate contract and submits one generation.

Widget-internal: load a page of the Marketing Studio v2 preset feed for a category.

Widget-internal: poll status and results of submitted Marketing Studio v2 jobs.

Confirm file uploads after using media_upload's upload_url method. Call this only after every curl PUT returned HTTP 200. Supports confirming multiple uploads at once via media_ids.

Import an HTTPS image, video, or audio URL into Higgsfield storage and return a confirmed media_id. Use this before generate_image/generate_video when the user provides a web media URL; generation medias should receive the returned media_id, not the original URL. Max URL payload: 50 MB.

Upload media for use in generation, or general files (documents, archives, code) for sharing. Returns presigned URLs for clients that can upload bytes themselves; run the generated curl commands or PUT the bytes to each upload_url, then call media_confirm. The media type is inferred from the filename extension: image/video/audio extensions become generation inputs; other whitelisted extensions (pdf, zip, tar, docx, csv, code files, …) are uploaded as general files and return a permanent URL, but cannot be used as generation inputs. General files are the agent's own upload path — the widget does not accept them, so upload the bytes to upload_url yourself (e.g. from a code execution environment). Supports batch uploads via files[]. Do not use this for user-provided local image/video/audio in Claude Apps UI-capable clients; call media_upload_widget instead so the user chooses the file in the Higgsfield widget and the browser uploads it directly.

Required local-media intake for Higgsfield in Apps UI-capable clients. Call this immediately as the only tool in the turn when the user refers to an attached/local photo, image, video, or audio but the prompt has no confirmed media_id yet. Do not inspect /mnt/user-data/uploads, run shell/sandbox commands, or ask the user to attach the file in Claude chat; remote MCP tools cannot read chat attachments. This widget is the upload surface: the user re-selects one or more files in the browser, the browser uploads them directly to Higgsfield storage, the widget confirms them, then sends the confirmed media_id/media_ids back to Claude for the next generation or analysis tool call. Use type auto with multiple enabled when one request needs mixed media, such as reference images plus a driving video; one video and one audio file may be combined with multiple images. The widget accepts media only; for general files (archives, documents, code) use media_upload instead and upload the bytes to the presigned upload_url yourself.

Find generation models. Use recommend with goal + input context; use get for model constraints. Items carry supports_unlim when the model accepts free-trial unlimited generations; the top-level unlim block says whether the caller can spend them right now, and the trailing 'Unlim configs' text lists the configurations their allowance actually covers.

Animate an existing character image with the motion and camera movement from a reference video using Kling 3.0 Motion Control. Use this when the user asks to recast, puppeteer, transfer motion, or make a character follow a driving clip. Pass image_id for the character still and motion_video_id for the reference motion video; each can be a confirmed uploaded media_id or a completed generation job_id. This tool does not use prompt or count; the scene prompt and background setup are handled automatically. resolution controls output quality, and scene_control chooses whether the background is based on the image or the video.

Expand or uncrop an existing image by outpainting beyond the original frame while preserving the source content. Use this when the user asks to extend the background, make an image wider or taller, change the canvas shape, or fill new edges around an image. Pass image_id for the source image and aspect_ratio for the target canvas. Optional width and height can be provided together; otherwise they default from aspect_ratio. This tool does not use prompt or count. Set params.get_cost=true to estimate credits without submitting a job.

Enter the website in the current Higgsfield app contest, together with the social-media links promoting it. A website not yet PUBLISHED to the community feed is published automatically by the entry — no need to call publish_website first. The website DOES need a live production deploy (deploy_website), else the entry is rejected. BEFORE entering, make sure the page metadata in app/src/app-meta.json is filled with real values (og_title etc.) — the auto-publish lists the website on the feed and an empty og_title makes it INVISIBLE there. Pass one or more urls, each a social-media link (YouTube, X/Twitter, Instagram, or TikTok); any other host is rejected. There is a single active contest, so no contest id is needed. Calling again for the same website OVERWRITES its urls (use it to fix or add links), it does not create a second entry.

Turn YouTube videos into ready-to-share clips. This is a long-running job and can take up to 30+ minutes. Before starting, ask the user how many clips they want, which clip aspect ratio to use, and which subtitle font they prefer.

Show recent clipping jobs.

Check clip creation progress.

Show available Higgsfield presets for image-to-video generation. Returns preset ids, names, previews, and descriptions.

Publish the website: lists the website's CURRENT LIVE production deploy on the Higgsfield community feed ('show in feed'), where other users can discover it. This does NOT deploy — deploy_website (which every build flow already runs) must have shipped the latest changes first; publishing with undeployed changes lists the OLD live build, and re-publishing does not re-deploy. BEFORE publishing, the page metadata in app/src/app-meta.json MUST be filled with real values — og_title, og_description, favicon_url, og_image_url — the feed card renders from them (read fresh from the pushed repo at publish time) and a website with an empty og_title is INVISIBLE on the feed; the live page's own head tags are baked at build time, so deploy AFTER changing them. Also OFFER the user a cover video for the card (og_video_url) — ask their permission first (video generation costs credits), never generate it unprompted. Commit and git push the metadata (and all other changes), then deploy, BEFORE calling this. Publish when the user asks to publish / share / go live on the feed, OR when they opted in to publishing at the start of the build — in that case publish automatically once the site is deployed with its metadata filled, without waiting to be asked again. For a plain deploy without a feed listing use deploy_website instead. EXCEPTION: a website whose production was never deployed (or was taken down by unpublish) falls back to deploying first — that returns status 'pending' while CI runs and the website is listed automatically once the deploy succeeds (check with website_status).

Expand or reframe an existing video to a new aspect ratio while preserving the source content. Use this when the user asks to make a video vertical, horizontal, square, wider, taller, or fill new edges around a video. Pass medias with exactly one source video and aspect_ratio for the target canvas. Optional image references can guide the filled area; optional start_image can pin the first frame when the user provides a first-frame anchor. For source videos over 15 seconds, pass duration_seconds and resolution and use only the source video. This tool does not use prompt or count. Set params.get_cost=true to estimate credits without submitting a job.

Remove or cut out the background from an existing image or video. Use this when the user asks for background removal, a transparent background, an isolated subject, a clean cutout, or a subject-only asset. Pass media_id for the source media and media_type as image or video; the matching background remover is selected automatically. This tool does not use prompt, count, or style parameters.

Rename the website's SUBDOMAIN (the slug in its public URL). The site is re-deployed under the new subdomain and the OLD subdomain STOPS WORKING — anyone holding the old URL must be given the new one. Storage (database, files, config) and the code repo are KEPT; only the public address changes. Runs a full re-deploy and can take a couple of minutes; returns once the site is live at the new URL. Fails if the new subdomain is already taken or reserved, or if a deploy is already in flight — pick another subdomain and retry.

Resolve a explainer video style preset (from get_explainer_presets) into a style reference media_id: the backend imports the preset's style image into the user's media storage. Pass the returned media_id as the style reference image in generation calls for every scene of the explainer.

Confirm the user has rights to the content of an `ip_detected` generation and flip its status to `completed`. Backend accepts only seedance-family jobs (cs_3_0, seedance_2_0, ms_video, etc) and only while the job is still in `ip_detected` state. Returns the updated generation. Used by the job-list widget's Reveal button after the user accepts the rights confirmation modal.

Execute a shell command in a remote Higgsfield cloud Linux sandbox — NOT your local machine or the client's own shell. Whenever a task needs shell tooling (ffmpeg, image/file conversion, scripting), use this tool, never a built-in or local bash/shell tool: only this sandbox has the media toolchain preinstalled and can reach the user's Higgsfield media. Preinstalled: ffmpeg/ffprobe, ImageMagick, sox, python3 with Pillow and faster-whisper, node/npm/npx, sharp-cli, Playwright with headless Chromium, caption fonts (Metropolis, Montserrat), zip/unzip, git, curl, jq. Use it for media processing (trim, convert, overlay, concat with ffmpeg), image manipulation, file conversion, scripting, and packaging that dedicated tools don't cover. The sandbox is isolated per user and is discarded ~10 seconds after a call finishes, so files in /home/user only survive between back-to-back calls — chain multi-step work into a single command (&&) and export results before finishing, or expect to re-download inputs. It has internet access: bring files in with curl from media URLs (media_import_url or generation results). For an output created here, call media_upload BEFORE starting the producing command, then append `curl -f -X PUT --upload-file <file> '<upload_url>'` to that SAME command so the ephemeral file is uploaded before it exits; call media_confirm only after HTTP 200. Never pass a sandbox path to media_upload_and_confirm: that tool accepts only client attachments. Commands run in /home/user and time out after timeout_seconds (default 60, max 120); for longer work (large renders, installs) set background:true and poll the returned log/status files with later sandbox_exec calls. Background work receives a 15-minute sandbox lease, and shorter poll calls never reduce its remaining lifetime. Set restart:true to discard the sandbox and start clean. Workflow bundle scripts are already installed in every sandbox under $HF_WORKFLOWS (/home/user/.higgsfield/workflows), laid out as $HF_WORKFLOWS/<workflow>/scripts/... — run them straight from there (they survive restart:true), and never paste script contents into the command.

Create a new private 3D Jutsu project for the authenticated user. Supply a descriptive name; the service assigns ownership and the project ID. Use this when the user requests a new project or a new standalone scene. For an existing scene, use scene_builder_3d_list_projects instead. Pass the returned projectId explicitly to subsequent tools; creation does not set a global active project. Call scene_builder_3d_get_project, then scene_builder_3d_query_python to inspect the initial scene and obtain guards before scene_builder_3d_run_python or scene_builder_3d_import_asset. Creation alone does not produce a committed GLB: scene_builder_3d_show_scene becomes available after the first successful edit or import. This call is not idempotent. If the response is interrupted or uncertain, use scene_builder_3d_list_projects to find the new project before retrying; repeating creation can create a duplicate. Finish each completed scene creation, edit, or import task by calling `scene_builder_3d_show_scene` once as the final 3D Jutsu tool call before your final reply, after mutations have settled and verification is complete. Pass the same `projectId` and the exact committed `revision` of the final result: `revisionAfter` from the final successful mutation, or the settled `project.revision` after an import. Do not guess the revision. This shows the result to the user; a text summary or download link alone does not finish scene delivery.

Resolve a short-lived download for an image or video that a successful Python operation published through `artifacts`. Use the artifact ID and operation ID or revision returned by that operation; do not invent IDs or filesystem paths. Query artifacts can be operation-scoped without a new committed revision. This retrieves an existing artifact and does not render one. To create a preview, render and publish it with `scene_builder_3d_query_python` or `scene_builder_3d_run_python`. Inspect the returned image using the client's image capability before judging framing, lighting, materials, and geometry. If the client cannot inspect it, describe that limitation rather than claiming a visual check passed. Finish each completed scene creation, edit, or import task by calling `scene_builder_3d_show_scene` once as the final 3D Jutsu tool call before your final reply, after mutations have settled and verification is complete. Pass the same `projectId` and the exact committed `revision` of the final result: `revisionAfter` from the final successful mutation, or the settled `project.revision` after an import. Do not guess the revision. This shows the result to the user; a text summary or download link alone does not finish scene delivery.

Resolve a short-lived download for the current committed editable Blender file, or a specified historical revision. This retrieves the existing scene; it does not create a revision or render a preview. Settle any active mutation with `scene_builder_3d_get_operation` first when you need its result. Use `scene_builder_3d_get_glb` for portable model delivery and `scene_builder_3d_get_artifact` for published images or videos.

For an interactive scene preview use `scene_builder_3d_show_scene`. Resolve a short-lived download for the current committed GLB, or a specified historical revision. This retrieves an existing export; it does not run Blender, render, or create a new export. Settle any active mutation with `scene_builder_3d_get_operation` first if you need its result. Use GLB for portable scene delivery, and `scene_builder_3d_get_blend` for the editable Blender source. Procedural shading, world lighting, and some Blender features may differ in GLB. A successful download does not establish that the exported scene looks correct.

Read or wait for a submitted Python operation in this project. Only `succeeded`, `failed`, `timed_out`, and `expired` are terminal; all other statuses require another poll using the same project and operation IDs. An HTTP success or a wait timeout does not mean the operation finished. Do not submit another mutation while one is active. On success, use the returned revision and scene sequence for subsequent work; obtain published images and videos through `scene_builder_3d_get_artifact`. On failure, inspect the error and re-read project state before deciding to retry. Do not blindly resubmit a failed operation under another ID. Finish each completed scene creation, edit, or import task by calling `scene_builder_3d_show_scene` once as the final 3D Jutsu tool call before your final reply, after mutations have settled and verification is complete. Pass the same `projectId` and the exact committed `revision` of the final result: `revisionAfter` from the final successful mutation, or the settled `project.revision` after an import. Do not guess the revision. This shows the result to the user; a text summary or download link alone does not finish scene delivery.

Read a 3D Jutsu project's current revision, scene sequence, active operation, and committed artifacts. Obtain `projectId` from `scene_builder_3d_list_projects`, `scene_builder_3d_create_project`, or an explicit user selection. An authorized project with `exists: false` is a valid empty scene at revision 0. Use `scene_builder_3d_query_python` to inspect actual objects, dimensions, materials, cameras, and lights before editing. If an operation is active, use `scene_builder_3d_get_operation` to settle it before another mutation. Scene edits use `scene_builder_3d_run_python` with the exact revision and scene sequence inspected; never guess these guards. Finish each completed scene creation, edit, or import task by calling `scene_builder_3d_show_scene` once as the final 3D Jutsu tool call before your final reply, after mutations have settled and verification is complete. Pass the same `projectId` and the exact committed `revision` of the final result: `revisionAfter` from the final successful mutation, or the settled `project.revision` after an import. Do not guess the revision. This shows the result to the user; a text summary or download link alone does not finish scene delivery.