Clipcat-ai/clipcat-skill

What it is

Clipcat is a software tool that turns an AI assistant (like Claude Code, Codex, or Cursor) into a TikTok Shop video producer — all through command-line commands the AI can run. It helps you find viral TikTok videos, research products and shops, turn product photos into AI-generated videos, write selling scripts based on real best-selling videos, and publish finished videos to TikTok.

It ships in two parts: a small command-line program called clipcat, and a SKILL.md file that tells an AI agent which commands are available and how to use them.

How it works

The AI agent reads the SKILL.md file to learn what it can do. Then the agent runs clipcat commands and reads the JSON output the program returns. Video and image generation tasks are submitted immediately, and the agent checks back later until they finish. OpenClaw installs the skill automatically from the manifest; other agents install the CLI manually.

What it can do

• TikTok shop data research — Query six types of entities: creators, products, shops, videos, live rooms, plus keyword and image search. Includes leaderboards, trend data, reviews, and relationships between entities.
• Viral selling-prompt generator — Search a library of 5,000+ real AI selling videos that actually sold on TikTok Shop (across 15 markets and 30 categories). Each video comes with the exact prompt that produced it, plus its real sales figures. The tool finds ones relevant to your product and rewrites the best match into a ready-to-shoot prompt, citing the original video.
• Video analysis — Extract scripts, scenes, hooks, and music from TikTok or Douyin videos.
• Viral replication — Recreate a proven viral video structure with your own product, based on a TikTok/Douyin link, a direct video URL, or a local file up to 100MB.
• Product-to-video — Turn product images into UGC-style TikTok videos.
• AI image generation — Generate e-commerce images from text prompts using GPT Image 2 / 2.5, optionally with up to 5 reference images.
• Super-resolution — Upscale generated videos to 720p, 1080p, or 2K.
• Reusable characters — Keep the same on-screen character across videos with --character-id.
• Publish to TikTok — Post a finished video now, schedule it up to 30 days ahead, or send it to drafts. Publishing costs no credits.
• Video download — Download TikTok or Douyin videos via the Clipcat API.

Installation

1. Install the CLI: OpenClaw does it automatically; otherwise run a one-line install command (curl -fsSL https://clipcat.ai/cli | bash).
2. Get an API key from the Clipcat website.
3. Configure the key on the machine the agent runs on with clipcat config --api-key your_key --base-url https://clipcat.ai, or set the CLIPCAT_API_KEY environment variable.

How you use it

You talk to your AI agent in natural language. Examples from the README: "Search viral TikTok videos about lip gloss this week," "Replicate this TikTok video with my product images," or "Publish this video to my TikTok account tomorrow at 8pm."

The agent figures out the right commands and shows you the plan before spending anything.

Key safeguards

• Before a video task consumes credits, the agent submits once for free to show you a confirmation checklist (model, duration, resolution, full prompt, credit cost) and waits for your explicit "yes."
• Nothing goes live on your TikTok account without your approval — the agent shows the full post checklist (account, caption, visibility, comments, schedule) first.
• Publishing needs a one-time account authorization and only works with watermark-free videos.
• You should not retry tasks manually; Clipcat already handles retries.
• Always include complete TikTok/Douyin URLs, including signed parameters.

Video models

There are 12 models, each with its own duration and resolution options. A few examples: grok_imagine (default trial model, 9:16 only), veo3.1fast (Google's Veo 3.1), seedance2_5 (ByteDance, clips up to 30s), and minimax_h3 (MiniMax H3, supports 768p and 2K). Trial models are free for everyone; standard models require a paid plan. The live listing (via clipcat models) is the authority on which models exist and what they cost per combination.

Languages and regions

Supports 11 languages (English, Chinese, French, German, Malay, Vietnamese, Thai, Japanese, Korean, Indonesian, Filipino, Spanish) and 15 regions: US, GB, DE, ES, FR, IT, JP, MX, BR, ID, MY, PH, SG, TH, VN. All monetary values in e-commerce data are converted to USD regardless of region.

Repository layout

Follows the Agent Skills convention — one skill folder containing a SKILL.md manifest:

clipcat-skill/
├── README.md
├── README_ZH.md            # Chinese version
├── docs/SKILL_ZH.md        # Chinese skill manifest
└── skills/clipcat/SKILL.md # the manifest agents load

More info

Homepage with the latest guide and examples: https://clipcat.ai. Command reference: skills/clipcat/SKILL.md in the repo.