20 MCP tools for AI UGC video production. Search viral formats, generate avatars, create videos, stitch clips, generate captions. Full video ad pipeline from Claude, Cursor, or ChatGPT.
Scored across 21 tools
Most tools have clearly distinct roles, and descriptions go out of their way to separate the confusable ones (generate_image vs face_swap, clean_image only after face_swap, generate_video vs get_render_status). The hook-related cluster (search_viral_videos, scrape_video, analyze_hook, remix_hook) is where an agent could still hesitate, since several of them operate on the same concept of a 'viral hook' and the ordering between them is only implied.
Every tool follows a strict snake_case verb_noun pattern (analyze_hook, create_product, get_render_status, list_assets, search_viral_videos). No camelCase, no vague bare verbs, and the verb choices (create/get/list/generate/scrape/stitch) map cleanly onto their action types.
21 tools is on the heavy side, but this server models a multi-stage video production pipeline (product setup, avatar/image generation, hook analysis, stitching, captions, credits), so most tools earn their place. A few peripheral entries like get_instructions and enhance_prompt could plausibly be folded into guidance rather than separate endpoints.
The creation and generation lifecycle is well covered, but retrieval and cleanup are thin: products and assets can be created and listed but never fetched individually or deleted, and there is no tool to publish or schedule the finished video even though create_content_plan produces a posting schedule. These gaps will force agents into dead ends when a user asks to remove an asset or actually post the result.
Analyze hook frame(s) using AI vision to extract hook text and generate a video prompt. Costs 1 credit.
Inputs: firstFrameUrl (required), existingPrompt, additionalFrameUrls.
Check current credit balance and subscription status
Inputs: no named parameters.
Remove text overlays and watermarks from an image. Only use AFTER face_swap if the result still has visible text - face_swap already tries to remove text by default. Costs 5 credits.
Inputs: imageUrl (required).
Generate a 3-week content plan with daily posting schedule, hooks, and platform rotation
Inputs: productId.
Create a new product/brand
Inputs: name (required), url, description.
Get a pre-signed URL to upload a file (avatar, demo video, music, etc). Creates an Asset record.
Inputs: type (required), fileName (required), title, fileSize, +2 more.
Use Claude to improve a rough prompt into a detailed, high-quality generation prompt. Costs 1 credit.
Inputs: type (required), prompt (required).
Swap your avatar's face onto a REAL hook frame from a viral video. Costs 5 credits. ALWAYS use sourceImageUrl (the real viral frame) + faceImageUrl (the avatar). The 'prompt' field controls HOW the swap is done (default is optimized, only override if user gives custom instructions). NEVER put user instructions in specPrompt - that destroys the real frame.
Inputs: faceImageUrl (required), prompt, specPrompt, sourceImageUrl.
Generate an AI avatar image from a text prompt. Always saved as a persistent asset. Costs 5 credits.
Inputs: prompt (required), model, avatarName, aspectRatio.
Generate an image from a text prompt only (scenes, backgrounds, product shots). Do NOT use this for putting avatars on hooks - use face_swap instead. Costs 5 credits.
Inputs: prompt (required), model, resolution, aspectRatio, +1 more.
Generate a video from a still image using AI. Models: draft (10 credits, ~30s, fast preview), standard (25 credits, ~1-2min, good quality), best (30 credits, ~2-3min, highest quality). Default duration is 5 seconds. If this tool times out, use get_render_status to check if the video was saved. IMPORTANT: The image should have a good video generation prompt — use enhance_prompt first if the prompt is generic.
Inputs: imageUrl (required), model, title, prompt, +2 more.
Get YuzuGrow usage instructions including YOLO mode. Call this when the user asks for help, says 'YOLO mode', or when you need guidance on the tool flow.
Inputs: no named parameters.
Check if a recently requested generation has completed by looking for new assets. Use after generate_video if it times out.
Inputs: type, limit, since.
List uploaded assets (avatars, videos, images, music). Returns signed URLs.
Inputs: type, limit, search.
List all products/brands for this account
Inputs: no named parameters.
Remix a viral hook text for a different product using AI. IMPORTANT: Never name competitors (Duolingo, Babbel, etc.) in the remixed hook. Use generic terms like 'language apps' or 'other apps' instead.
Inputs: originalHook (required), productDescription (required).
Scrape a video URL to extract the first frame and metadata. Only stores the first frame (for face_swap). No video clips or extra frames saved. Free (0 credits). Use this when a database result has no cleanThumbnailUrl, or when the user brings a new URL.
Inputs: url (required).
Search the curated viral video database for hook inspiration
Inputs: limit, search, emotion, category, +1 more.
Assemble hook + demo clips into a final vertical video with text overlay. Costs 2 credits. IMPORTANT: Before stitching, ask the user whether they want the text overlay to appear on the demo section too or only on the hook.
Inputs: clips (required), projectName (required), fontSize, hookText, +7 more.
Generate 3 social media caption suggestions for a UGC video
Inputs: hookText, productContext.
Update an existing product/brand
Inputs: id (required), url, name, description.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Add one secure layer between your agents and this server.
