Tool search 164,478 tools · 10,067 live servers

Assemble a designed-looking promo/intro/product/data video from the Clipkit pattern library: give an ordered list of SCENES and the words, and it bakes in the camera, glass, lighting, motion blur, timing, and layout, then returns an editor link. This is a FAST option when a conventional promo structure fits — it is NOT the only way to make a video and NOT a default; for anything specific or original, author the JSON yourself and call set_project (the full creative range). When you do use this, MIX scene types to fit the brief and vary the structure — a video can be a single kinetic headline, three title cards, a showcase montage, or a data explainer; you do NOT need a hero or a cta. Scene types: hero (glass-orb logo reveal: wordmark, tagline?), kinetic (letter-fly headline: text, subtitle?), showcase (a screenshot tilted in 3D: screenshot URL), title (full-frame title card: headline, kicker?, subtitle?), cta (closing card with a glass button: wordmark, tagline?, cta), stats (hero numbers: stats[{label,current,previous?}], title?), bars (bar chart: bars[{label,value,previous?}], title?), ranking (top-N list: items[{label,value}], title?), pie (pie cards: cards[{label,value,total,previous?}], title?). The data scenes (stats/bars/ranking/pie) look best with theme "mux".

Use this when the user asks to screenshot, capture, or take a picture of a webpage/URL, or to render raw HTML or Markdown to an image or PDF. Do NOT use to get a reusable hosted image URL (use rendex_render_link) or to make a branded multi-format document (use render_artifact). Captures a screenshot or PDF of any webpage, raw HTML, or Markdown. Supports full-page capture, dark mode, ad blocking, custom viewports, CSS/JS injection, cookie/header injection, PDF output, HTML and Markdown rendering, and progressive fallback for heavy sites. Returns partial renders on timeout by default (bestAttempt mode). Costs 1 render credit per call. Cookie/header injection requires Starter+; geo-targeting requires Pro+.

Use this when the user asks to read, extract, get the text/content/article of, or summarize a webpage/URL. Do NOT use for a visual screenshot (use rendex_screenshot). Extracts clean reader-mode content from any webpage as Markdown, JSON, or HTML. Runs the same Chromium render pass as a screenshot, so it captures content after JavaScript runs — handles SPAs that fetch-only readers miss. Strips nav, ads, and boilerplate, returning the article body plus title, byline, and excerpt. Great for feeding page content to an LLM, summarization, or RAG ingestion. Costs 1 render credit per call.

Use this when the user wants a reusable, HOSTED image URL — an og:image or a link to embed in an <img> tag — rather than the image bytes. Do NOT use for a one-off inline screenshot (use rendex_screenshot). Renders a URL, raw HTML, or Markdown and gives back a signed, hosted, edge-cached image URL instead of the bytes — ideal for dynamic OG images: drop the URL into <meta property="og:image"> or an <img> tag and Rendex serves a cached copy on every share. Takes the same options as rendex_screenshot, plus an optional expiresIn. Returns { url, expiresAt, format, cacheTtl } as JSON. Costs 1 render credit per fresh render; cached repeat hits don't re-charge.

Use this when the user asks to make or create a branded report, invoice, summary, release notes, or document — or to 'turn this Markdown/HTML into a PDF and PNG' (optionally with a logo or accent color). Do NOT use to screenshot an existing URL (use rendex_screenshot). Turns Markdown or HTML into a branded, downloadable artifact — a PDF, a PNG, and a hosted share page — in one call. Apply a logo, accentColor, font, header, and footer. Returns hosted URLs { pdfUrl, pngUrl, shareUrl, expiresAt }. Each requested format costs 1 render credit.

Use this when the user asks to monitor, watch, or track a webpage for changes, or to be alerted/notified when a page changes. Do NOT use for a one-time capture (use rendex_screenshot). Creates a Rendex Watch — monitors a URL on a schedule and notifies when it changes (real-Chrome visual diff with a highlighted overlay, an extracted-text diff, or both). An active watch captures its baseline immediately. Returns the created watch as JSON.

[AI Search via Google SERP] Scrape publicly-available Google search results (data source: Google; use must comply with Google Terms of Service) with top AI Overview, organic results, and related searches. Two modes: overview (standard SERP) / ai_mode (immersive multi-turn conversational search). Use when: user says "Google for me" / "external demand" / "what do people say about X" / "Reddit/Quora pain points" / "will my content be cited in AI search" / "find user complaints for keyword X"; "consumer voice" step in scouting SOPs; verifying whether a new product concept has off-Amazon demand; **see which Google Shopping ads competitors run / their ad landing pages** (the sponsered block). Don't use: for on-Amazon search (use search_amazon); when only the trend curve matters (use keyword_trends — cheaper and tighter). Returns: data.{ results_num, ai_overview, json.items[ { type:'ai_overview', items:[{content:[...], references:[{title,url,domain}]}] }, { type:'organic', items:[{title,url,text}] }, { type:'related_searches', items:[...] }, { type:'sponsered', items:[{type:'result', url, position:'top'|'bottom', title_of_page, title_above_url}] } ], screenshot, taskId }. ⚠️ The ad block's upstream type is literally spelled 'sponsered' (missing an o — not a typo on our side; match it verbatim, do NOT look for 'sponsored') — it carries Google ad (shopping + text) landing-page url, title (title_of_page), and displayed brand domain (title_above_url). **position** marks whether the ad appears at the top ('top') or bottom ('bottom') of the page — top ads carry higher exposure weight. Pair with: ↑ query inferred from user; in 'ai_mode' pass followups[1..5] for multi-turn; ↓ ai_overview.references[].url for authoritative external sources, organic items for content-competition analysis, sponsered[].url + title_above_url for competitors' paid landing pages and brands, split by position into top/bottom ad slots. Cost: ~2 points/call, ~30s (**slow** — Google AI render time). Tips: prefer overview for single queries (cheaper); use ai_mode only when you need decomposed multi-turn investigation. Followups > 5 visibly slow down responses.

[Amazon Rufus AI conversational recommendations] Ask Amazon's AI shopping assistant Rufus in natural language, get grouped structured product recommendations + Rufus text reply + follow-up questions. Use when: user says "ask Amazon AI X" / "Rufus recommendations" / "find products conversationally" / "products for a scene (gifting / camping / moving)" / "open-ended sourcing" / "I have no keyword, just a scenario". Don't use: when you already have a clear keyword and want SERP (use search_amazon); category bestseller ranks (use list_bestsellers); single-ASIN detail (use get_amazon_product); Google-side AI search (use ai_search). Returns: data.json[{ prompt, content, products[{ title, items[{ asin,url,title,cover,score,ratingsCount,price,originalPrice,describe }] }], follow_up_questions[], screenshot }] + top-level taskId / url / screenshot. Note: follow_up_questions is snake_case (passed through from backend verbatim). Pair with: ↓ feed asin into get_amazon_product / get_amazon_reviews for deep-dive; follow_up_questions can seed the next round's prompts for multi-turn exploration. Cost: **6 points PER PROMPT** (billed by prompts count, NOT a flat 6 per call; N prompts = N×6 points). ⚠️ **Slow tool**: **strongly prefer sending exactly 1 prompt per call**. A single prompt typically takes **60–90s** (Rufus generates the conversation live — far slower than a normal scrape); multiple prompts add up linearly and **can exceed 200s**, costing both time and points. Treat this as a long-running call: set your MCP client's per-tool-call timeout to **≥120s** (many clients default to a 60s silent timeout that aborts before this tool returns, making the agent wrongly report it as "unavailable"), and do NOT retry or fire concurrent duplicate calls just because it didn't return instantly. When you send a progressToken, the server emits a progress heartbeat every 15s to keep spec-compliant clients' timers alive. For several needs, make several single-prompt calls rather than batching them.

Get the article content as markdown. Screenshots appear as <SCREENSHOT {id}> placeholders. Returns the markdown text and a list of screenshot IDs.

Update article text content from markdown. Use <SCREENSHOT {id}> placeholders to position screenshots. Screenshots in the markdown are kept. Screenshots removed from markdown are deleted. Order in markdown defines final order. Get the current article first with get_article, modify the markdown, then call this.

Add a screenshot or GIF to the article. Returns a screenshot_id to place in the article via update_article using `<SCREENSHOT screenshot_id>`. - kind="image": pass `timestamp` to capture a frame from the project's base video, OR pass `mcp_upload_id` (from the upload tool) to attach an uploaded image. One of the two is required. - kind="gif": pass `timestamp` (start) and `duration` (seconds). Captures a short GIF clip from the base video. Requires the project to have a video clip.

Update an article image/GIF (screenshot), or inspect its current state. - Provide one or more of `zoom_factor`, `crop`, `shapes` to change metadata. - Pass `render=true` (with no other updates) to fetch the current details without mutating. To DELETE an article image/GIF: call `update_article` with markdown that omits the `<SCREENSHOT id>` reference. Unreferenced media is pruned server-side on save. There is no separate delete tool — the markdown IS the source of truth for what's present.

Create a new clueprint (reusable video template). A clueprint is a structured file directory with design notes, composition rules, and screenshots that can be used to generate videos. Use human-readable folder categories: design/ (colors, typography, backgrounds, animation), rules/ (composition, voiceover, editing), screenshots/ (rendered clip previews). See the "Creating Clueprints" section in the system instructions for the full structure. If source_guide_ids are provided, guide data is automatically extracted clip-by-clip into the clueprint as reference material.

Evaluate a design description against UX principles. Returns relevant principles, potential violations, and improvement suggestions. Screenshot pixel-diff is disabled on the hosted (remote) endpoint (unbounded image decode). Omit 'before_screenshot'/'after_screenshot' and pass a 'description' to evaluate the design against UX principles statelessly.

Audit a rendered iOS screen from a view-hierarchy/accessibility snapshot (and optional screenshot). Alias of audit_screen with platform:"ios". Call with no arguments for the expected snapshot shape. Call with {elements:[{label,rect:{x,y,w,h},role,fontPt,fgColor,bgColor}],viewport:{w,h}} to score 44×44pt touch targets, contrast (with iOS secondaryLabel/tertiaryLabel treated as platform-standard — warn not fail), and visual rhythm (alignment, gap consistency, optical balance) in points. Same return shape as audit_page. Taste-profile bindings are not available on the hosted (remote) endpoint (per-user state is a later phase). Omit 'project'/'profile' to run the stateless iOS screen audit.

Identify which font is used in an image. Powered by our OWN CNN embedding model, trained on the jinero font catalog — it matches fonts by visual shape/style, so it needs NO OCR and NO text (works for Latin and Cyrillic). Send a tight crop of one line of text as either image_url (public URL) or image_base64 (base64/data-URI, e.g. a local screenshot). The image is processed in memory and deleted immediately — never stored. Returns the most visually similar font families with scores. Typically 0.3-1.7 s (an OCR-assisted rerank of top candidates engages when retrieval is uncertain). Rate limit: 5 calls/min per IP (model inference is compute-heavy).

Extract a dominant-color palette from an image. Send either a public image_url or image_base64 (base64/data-URI, e.g. a local screenshot). The image is processed in memory and never stored. Each color comes back with its hex, rgb, hsl, share of the image, and the closest human color name — a named palette in one call. Rate limit: 10 calls/min per IP.

Capture a screenshot of a web page and return it as a JPEG image. 입력한 URL의 화면을 캡처해 JPEG 이미지로 반환합니다. [호출당 5포인트]

Use this when you need to resize an image from a URL or convert its format. Returns the resized image as base64 with metadata in JSON. Returns: 1. image (base64-encoded data) 2. originalWidth and originalHeight 3. newWidth and newHeight 4. format (png/jpeg/webp) 5. fileSizeBytes. Example output: {"image":"iVBORw0KGgo...","originalWidth":1920,"originalHeight":1080,"newWidth":800,"newHeight":450,"format":"webp","fileSizeBytes":24500} Use this FOR creating thumbnails, optimizing images for web, preparing social media images, resizing for email templates, and batch image processing. Do NOT use for OCR text extraction -- use media_extract_text_from_image instead. Do NOT use for QR code generation -- use utility_generate_qr_code instead. Do NOT use for screenshots -- use capture_screenshot instead.

CREATES a fake-text video: a phone chat conversation that types itself out over a gameplay or green-screen background, with every line read aloud in the voice you pick for that side. It is one of aicut's most-used short-form formats. YOU WRITE THE SCRIPT. There is no script tool and no writer behind this endpoint - the `messages` array you send IS the conversation that gets made, in that order. Write it yourself from what the user asked for, show it to them as plain text before spending anything, and change it until they like it. Iterating on the script costs nothing. A WORKED EXAMPLE of `messages` - a four-line chat where the phone's owner is on the right: `[{"text":"i saw what you posted","from":"left"},{"text":"and?","from":"right"},{"text":"you told me you were working","from":"left"},{"text":"i was. at his place.","from":"right"}]`. That is the whole format: an ordered list of bubbles, each with the line and the side. Write the conversation the way a real one runs - short lines, one thought each, a hook in the first two bubbles, and a last line worth screenshotting. AFTER: the response carries the job id, the price split (`generation_tokens` for the spoken audio, `render_tokens` for the video file) and `renders_automatically: true`. If it ALSO carries `start_confirmed: false`, the job exists but aicut never saw its start confirmed - do not create it again, watch that job id and tell the user it may need a retry if it has not moved in ten minutes. THAT LAST ONE MATTERS: unlike an AI Video Story there is no fire step and no render step - this one call makes the finished video. Wait for it with `wait_for_generation`; when it is terminal, `get_video` carries the file url. REFUSALS (the common ones, not all of them - always read the `code` you actually get): 400 = the script or a setting is not accepted, and the message says which (an empty line, a missing `from`, an unknown template or voice, or the whole chat reaching 6000 characters of dialogue, which is the ceiling - shorten it, aicut will not silently cut it for you). 402 = not enough tokens for the whole video; the body carries `required` and `balance`. 503 `fake_text_unavailable` = the video was not started and nothing was charged; retry the same call once. COST: quoted per character of dialogue, in two parts that are both charged - call with `estimate_only: true` first and state `estimated_tokens` (the total) to the user. A longer script costs more, AND SO DOES THE VOICE PROVIDER: ElevenLabs is about four times openai/polly per character. Change a provider and re-run `estimate_only`; never carry an older quote across that change. WHAT THIS TOOL DOES NOT OFFER, so you do not go looking: background music, specific gameplay clips, images or sound effects inside the chat, contact photos, promotions, a second chat in one video, and the `cantina` skin. Those live in the aicut web app - say so rather than approximating them in the script. (Cloned and custom voices ARE offered - see `voice_left` / `voice_right`.) THE WATERMARK is decided by the account's plan, not by this call: free accounts get the aicut mark on the video. Say so if the user asks; there is no argument that changes it. DELIVERY: hand the user ONE thing - the finished video. Do not re-list the script back at them after it is made. SPEND ETIQUETTE (the money grammar): in the webapp the priced button is the user's own finger; in chat YOUR tool call is not - so state the price IN THE SAME MESSAGE as the ask, and the user's explicit go is the button press. Never charge on inference: quoting is not asking, and after a price you wait for the yes. THIS APPLIES TO EVERY TOOL CARRYING THIS NOTE, including this one. A GO IS SCOPED TO ONE PURCHASE, AND IT MUST BE UNAMBIGUOUS. The user's instruction has to NAME the thing you are about to buy, or refer to it so plainly that it cannot mean anything else. A BARE AFFIRMATION - 'go', 'yes', 'ok', 'do it', 'just do it', 'sure' - counts ONLY when ALL THREE of these hold: the message immediately before it was YOUR priced ask for THAT EXACT action, nothing else was raised in between, and NOTHING THE USER ASKED FOR EARLIER IS STILL OUTSTANDING. That last one is the trap the others miss: if the user's OWN previous turn asked for something else - a refusal, a different scene, an edit, a redraw, a question - their 'just do it' may be answering THAT, and it is AMBIGUOUS even when your priced ask happens to be the last thing said in the thread. An ambiguous affirmation is not a go: ask WHICH one they mean and state that price again. WHEN IN DOUBT ABOUT WHAT A 'GO' REFERS TO, ASK. A wrong guess spends the user's money on something they never asked for, and nothing on this surface can undo it or give it back - asking costs one sentence. An episode's STAGES - cast portraits, episode create, fire, render - are each their own priced ask. A STANDING GO IS NOT UNLIMITED: 'just make it' or 'go ahead with the whole episode' authorizes the stages you PRICED IN THAT SAME MESSAGE, in the order you named them, and nothing beyond them - so do not re-ask per stage while it holds, and do not stretch it over a stage whose price the user never saw. IT EXPIRES THE MOMENT THE USER RAISES ANYTHING ELSE - a change, a question, a refusal, a redraw, a new idea - and after that the next stage needs its own priced ask. ONE STAGE IS NEVER COVERED BY A STANDING GO AT ALL: the FIRE (`fire_story_video`) is irreversible and the biggest single charge in the episode, so it always takes a go that NAMES firing, whatever was said earlier - see that tool's own note. A REDRAW IS NOT A STAGE: `regenerate_story_frame` and `regenerate_cast_portrait` are extra spends the user asks for one at a time, so state that price every time, even under a standing go. A standing go never carries to a different episode, and never to `generate_video`, `generate_image` or `generate_audio` - each of those is its own ask. ACCOUNT FOR YOUR OWN CALLS: if the user says something happened that you did not intend - a charge they did not expect, a step they did not ask for - RE-READ YOUR OWN TOOL CALLS IN THIS CONVERSATION before you answer, and tell them plainly which tools you called and when. NEVER SPECULATE ABOUT A CAUSE YOU CANNOT OBSERVE: not a button on an aicut card, not the user's own click, not their client. The aicut cards CANNOT SPEND - the only tools they ever call are the reads (`get_video` / `get_image` / `get_audio`), and their buttons either save a file or send a VISIBLE user turn into the chat - none of them calls a spending tool - so saying a card might have generated or charged something is false, not a hedge. (If a spend followed one of those visible turns, it was still YOUR call, and the honest answer names it.) If your call history disagrees with what you told the user, say what you actually called and let them correct you; do not invent an explanation that makes the two agree. IDEMPOTENCY: `idempotency_key` is optional and makes a retry safe. Set it on the FIRST call, not only on a retry - the job is addressed by the key, so a key added afterwards cannot find a job that was created without one. Reusing a key REPLAYS the job that key already created and returns it unchanged - even if you send a different prompt or different settings, and even after that job has finished. A key is therefore spent permanently. Do NOT reuse one to make another generation: two deliberate generations are two jobs and need two different keys (or none). ONE EXCEPTION, on `render_story_video`: replaying a key whose render FAILED answers 409 `render_failed` rather than replaying the failure, because a spent key stays spent - retry that one with a NEW key or with none. A KEY IS NOT SCOPED TO A TOOL: it addresses a job on the whole account, so reusing the key you gave `generate_video` on `generate_story_video` replays that first video instead of starting an episode. One key, one thing you made. (`render_story_video` is the one door that namespaces its own, which is why an episode's key can be reused on its render without colliding - but there is no reason to reuse it there either.) Never derive the key from the request body. You do NOT need to pass one to be safe against a duplicated delivery: aicut already derives a per-call key server-side, so a retry the transport makes on its own replays rather than charging twice. Pass your own only when YOU want to retry a call whose answer you never saw. OUTPUT: this returns JSON for you to read. When you report back to the user, give them the media URL plus a one-line summary. Do not paste the raw JSON, job ids, or internal field names into the conversation.

IS A RECOVERY PHRASE SITTING IN CLEARTEXT ON THIS MACHINE? Everything else here answers whether an ADDRESS is safe to pay; this answers whether the MACHINE is safe to hold a wallet, and a safe address on a compromised machine is worth nothing. "Self custody if you know how to keep your seedphrase safe" puts the whole condition in the sentence and nothing ships that checks it: an antivirus answers "do you have a known virus", which is a different question. This one is DECIDABLE rather than scored. A keyword scan drowns — abandon, able, about and absent are ordinary English and all four are BIP-39 words — but a mnemonic is a RUN of 12/15/18/21/24 consecutive words from a 2048-word list with a CHECKSUM in the last word, so a candidate is proven by arithmetic. Measured across 204 files and 1.6 MB of real prose and source: zero false confirmations. It NEVER outputs the phrase — file, line and word count only, because this output ends up in terminal buffers, logs and screenshots, and a scanner that prints the seed it found is a stealer with good intentions. Reports its own blind spots: no images, PDFs, password managers, browser storage or encrypted archives, so "nothing found" means nothing was found IN WHAT WAS READ. Read-only, no network, nothing is copied.

SECONDARY (user-driven UI): turn a screen's REFERENCE IMAGE into placed elements. Returns the image itself plus the recipe — the canvas resolution, the fraction→pixel conversion that stops coordinates landing wrong, and what is already on the screen so a second pass updates instead of duplicating. You look at the picture and send back propose_element / update_element with x/y/w/h; it all lands in the owner's Inbox. Use when the user wants the wireframe to match a screenshot or mockup they uploaded.

Fetch an image from a public URL and re-encode it smaller (WebP/AVIF/JPEG), optionally resizing to a target width. Returns the optimized image plus before/after byte sizes. Use this when an AI-generated or dropped-in asset (hero image, screenshot, illustration) is too large to ship.

Take a screenshot of a live web page by URL using headless Chrome. Returns a { FileUrl } (valid ~24h). BILLED. Api2Pdf API: POST /chrome/image/url.

Capture a screenshot of a web page (additive; CONSUMES API quota). Async by default: returns a `pending` screenshot to poll. Set synchronous:true to block and receive the finished screenshot via the sync host. API: POST /v2/screenshots.

Image to text: read the words out of a photo — receipts, screenshots, handwriting. A HelpySelf tool (helpyself.com).

A barcode scanner online — read a barcode or QR code from a photo or screenshot. A HelpySelf tool (helpyself.com).

Resizes a batch of 1+ images to a named real-world preset (App Store screenshots, Open Graph/social cards, Android launcher icon densities, a favicon set) or a custom width/height, using contain (whole picture, padded), cover (fills the frame, crops overflow) or stretch (distorts to fit) fitting. Each resized image is PNG-encoded (EXIF/ICC metadata stripped on re-encode, matching the source page) and every image is auto-rotated per its embedded EXIF orientation before resizing. All outputs are packaged into one uncompressed ZIP (PNG bytes are already compressed, so a second pass would not shrink them), foldered by preset (e.g. "social/photo-1200x630.png", or "android/mipmap-hdpi/icon.png" for the Android density preset, which names each size's folder after its density instead of suffixing the pixel size). This always returns as a resource_link (a batch ZIP is never small enough, or singular enough, to inline) -- fetch the link to get the archive. Per-file output dimensions and byte sizes are reported in the JSON result. Each input image (base64) is capped by this server's per-input byte limit.

Composite one or more app screenshots into a drawn device frame (rounded bezel, Dynamic Island on phone-shaped canvases) with a gradient background and optional headline/caption text, and export as PNG at any/all of the 4 official App Store Connect sizes. A blank page (image omitted) draws no device, for a title card or an outro. A device positioned past the edge of its page is drawn spilling onto the neighbouring page too, so a set can run one screenshot into the next. LIMITS — this runs on a small shared host, so an over-limit call is refused immediately, before any rendering, with a message saying how to split it: at most 10 pages at any one size, 2 at size:"all" (which renders every page 4 times over — a longer set needs one call per size); at most 12 rendered PNGs per call (pages x sizes); 6 MP per screenshot and 4 MB of screenshot bytes summed across the whole call (send JPEG, not PNG, for a multi-page set); and at most 40 text layers of 512 characters each.

Measured design CHANGE HISTORY for a live-decoded domain — the Decode Ledger. Token-level diffs between deep decodes over time: "radius 4px→8px", "primary hover #4032C8→#0A2540", "motion dominant 150ms→200ms", each dated. Use it to see how a product's design system is EVOLVING (no screenshot library can backfill this). site = a domain ("stripe.com") or product name. Returns first/last decode dates, decode_count and the dated change entries; empty history = measured, stable so far.

Has this product's design system CHANGED since it was decoded — and which of its numbers are safe to hard-code? Mozaika re-measures the most-referenced products from the live DOM every night and keeps a dated ledger, so this answers what a screenshot never can: • verdict "held" — nothing moved for N consecutive nights; the spec is still accurate. • verdict "shifted" — a token changed and the new value stuck (with the date and before/after). • verdict "unstable" — a value alternates between nights: a live A/B test, rotating content, or a page that renders differently each run. `do_not_hardcode` lists it. Call this BEFORE building against a cached spec, and before baking any measured value into a token file. Pairs with get_design_system(site) — that gives the spec, this gives its shelf life. Args: domain: e.g. "stripe.com", "linear.app" (bare domain, no scheme). Free.

The universal BUILD-KIT fetcher — the measured spec + code to reproduce a piece of UI. `recipe_type` selects which library across three families (all agent-ready through this one call): • COMPONENTS (decoded live from a real product's DOM — Mozaika's wedge): "Command Palette", "Dropdown Menu", "Dialog / Modal", "Login", "Data Table", "Onboarding Tour", "Navbar", "Logo Marquee", "Toast", "Date Picker", "Combobox" — returns the anatomy TREE (each node measured), the MOTION (open/close animation + easing a screenshot can't show), every STATE (empty/results/no-results/keyboard-selected), a webm of it running, and the design tokens. • EFFECTS (open-source WebGL hero backgrounds — Apache/MIT): "Hero Effect" → the shader's full config + fps + install command + license/NOTICE. • MOTION (open-source looping showcase templates — MIT): "Motion Showcase" → the template's full parameter surface + the exact Swiper/anime.js/Motion config + install. Use for "build a <thing> like <product>", e.g. get_recipe("Vercel", "Command Palette") or get_recipe("vanta", "Hero Effect"). A complete, uncopyable, measured build kit. Returns the available types if the requested one isn't found.

Re-pose an existing sprite into a new target pose while preserving the character, taking a source image (URL or base64), a pose name (or "Other" with a free-text description), and an optional n (1-4) for how many variations to produce. Only works with sprite image types (not icons, screenshots, etc.). The job result is an array of pose results, each containing the generated image url, the pose and description used, and a suggested motion_prompt tuned for that pose. Credits are charged only on success, scaled by the number of images generated. This is typically the first step before animating: call generatePose to set the character's pose, then feed the result (and its suggested motion_prompt) into animateSprite for the best animation quality; use rotateSprite instead when you want to change the camera angle rather than the pose. Pass an optional request_id to tag the results so you can locate them later via `GET /assets/sprites/results`. Requires an API key (user scope). Returns 202 with a job id immediately; poll `getApiJob` (pass `wait: 30`) until status is succeeded, then read its `result` field, which is exactly the response documented for this operation. Each account may have up to 50 generations queued or running at once; beyond that submissions return 429 (PENDING_JOBS_LIMIT) - wait for jobs to finish. Credits: This endpoint consumes 0.5 credits per result.

Re-render an existing sprite from a different camera viewpoint while keeping the same character and pose, taking a source image (URL or base64), a required camera_rotation azimuth (one of 0, 45, 90, 135, 180, -135, -90, -45 degrees), an optional camera_elevation (0, 30, or 60 degrees; omit to keep the current elevation), and an optional n (1-4) for the number of variations. Only works with sprite image types (not icons, screenshots, etc.). The job result is an array of rotate-sprite results, each with the generated image url and the camera_rotation and camera_elevation that were applied. Credits are charged only on success, scaled by the number of images generated. Use this to produce alternate view angles of a character; use generatePose instead to change the character's pose rather than the camera, and animateSprite or transferMotion to bring a sprite to life. Pass an optional request_id to tag the results so you can retrieve them later via `GET /assets/sprites/results`. Requires an API key (user scope). Returns 202 with a job id immediately; poll `getApiJob` (pass `wait: 30`) until status is succeeded, then read its `result` field, which is exactly the response documented for this operation. Each account may have up to 50 generations queued or running at once; beyond that submissions return 429 (PENDING_JOBS_LIMIT) - wait for jobs to finish. Credits: This endpoint consumes 0.5 credits per result.

Render HTML/CSS to a PNG image via headless Chromium. Use this when the user explicitly asks to render HTML/CSS, provides HTML, or requests a deterministic, pixel-controlled layout assembled with web code. Do not choose this tool for ordinary AI image generation or editing, for applying a theme reference image generatively, or when the user names an image model. A request for a social card, banner, poster, or OG image by itself is not enough to select this tool; use generate_image unless the user specifically asks for HTML/CSS rendering or deterministic code-based layout. **Tailwind CSS:** All Tailwind utility classes work out of the box — no CDN script or stylesheet needed. Just use classes like bg-blue-500, text-white, flex, rounded-xl, shadow-lg directly in your HTML. **Full CSS surface:** flexbox, grid, filter (blur, grayscale, hue-rotate, drop-shadow), mask-image, object-fit, transform, gradients, <style> blocks, class selectors, pseudo-elements, variable fonts (all weights 100-900 + italic). Write HTML like you would for a real browser. **Fonts:** Any web font works — this is a real browser. Load fonts directly in the HTML: `<link rel="stylesheet" href="...">`, `@import url(...)`, or `@font-face`. Works with Google Fonts, Bunny Fonts, Adobe Fonts, your own CDN, etc. For a custom/local font, pass it via inline_images and reference it from `@font-face { src: url('{{my_font}}') format('woff2'); }`. The renderer waits on `document.fonts.ready` before screenshotting, so whatever the page declares is what gets rendered. No default font is injected — be explicit. **Typical cold-render time:** ~1-3s (first render pays the browser cold-start; subsequent renders reuse the browser and complete in ~700ms-1.5s). Concurrency is capped server-side. For reusable code-based layouts, use create_template + render_template instead.

Start a temporary, email-free Pulse trial. Returns a short-lived private API key with enough banked tokens for one short URL-based v1 analysis. No Stripe Checkout session is created. Store the key in the client or a trusted local environment; do not repost it in public logs, screenshots, GitHub issues, shared chats, or directory examples.