OnVid Prompts
A free connector for Claude, ChatGPT and other MCP clients that writes, improves and finds prompts for AI video models: Veo, Sora, Kling, Seedance, Hailuo, Wan, Runway and others. It returns text only. It does not create videos, needs no account and stores nothing you send.
Server URL (Streamable HTTP, no authentication)
https://mcp.onvid.ai/promptsAdd it
- Claude (web, desktop, mobile): Customize → Connectors → Add custom connector, paste the URL above and save. No sign-in step.
- Claude Code:
claude mcp add --transport http onvid-prompts https://mcp.onvid.ai/prompts - ChatGPT: in developer mode, add a connector with the URL above and authentication set to none.
- Other clients (Cursor, VS Code, LM Studio): add a remote MCP server with the URL above.
Tools
All tools are read-only.
build_video_prompt: Turns a short idea into three ready-to-use prompts (action, camera move, lighting, sound, length, aspect ratio) plus a multi-shot version, worded for the model you pick. Handles animating your own photo and keeps spoken lines word for word.improve_prompt: Keeps your words, adds what's missing (camera, light, style, sound, length) and lists problems to fix, such as too many story beats for a short clip.suggest_prompt_ideas: Video ideas for a topic in formats that work on TikTok, Reels, YouTube and ads: POV, before and after, tiny world, talking pet, movie trailer and more.search_prompts: Searches a library of 80 video prompts in 16 categories, from cinematic scenes and product ads to dance, pets, anime, horror and holidays.get_prompt: The full text of one library prompt, filled in with your own subject and worded for your model.get_model_prompt_guide: How to write prompts for one model: structure, sound and dialogue, negative prompts, clip length and tips.list_prompt_categories: The categories, uses, styles, models and lengths the builder understands.
How prompts are worded per model
- Veo, Sora, Kling and Seedance get a sound line, and dialogue written the way each model reads it.
- Kling and Wan get a separate negative prompt; other models get a short "Avoid" line.
- Hailuo prompts start with a bracket camera command such as [Push in].
- Runway image-to-video prompts describe only the motion, worded positively.
- When you animate your own photo, every model is told to keep the person, product and background unchanged.
Try asking
- “Write a Veo prompt for a corgi in a raincoat saying "I did not eat the cake".”
- “Improve this prompt for Kling: a cat walks on the moon then dances then flies away.”
- “I'm animating a photo of my grandma in Runway. Give me three prompts.”
- “Give me six TikTok video ideas for my coffee shop.”
- “How should I write prompts for Seedance?”
Privacy and support
The connector needs no account and keeps no copy of your prompts. Our logs record only which tool was used and when. Details are in section 6 of the Privacy Policy. Questions or problems: [email protected] or the Support page.
Want to make the videos too? The OnVid studio connector creates them with your OnVid account.