Stivio for AI clients
Your next video,
from your AI workspace.
Connect Stivio through the Model Context Protocol. Give your AI client an image and your motion description, approve the credit cost, and get a video back.
Image-to-video and text-to-video · Your existing Stivio credits · MP4 results
Connect to Stivio
Use a client that supports remote MCP over Streamable HTTP and a custom Authorization header, such as Cursor. Add this server URL:
https://stivio.ai/api/mcp
Your MCP key
A key lets your AI client upload images, spend your Stivio credits on videos, and read your balance and video results. It expires after 30 days. You can have up to five active keys. Keep keys private and only give them to clients you trust.
Loading account…
Client configuration
Replace YOUR_STIVIO_KEY with your key in your client's private MCP settings. This example uses Cursor's mcpServers format. Other clients may use a different configuration format.
{
"mcpServers": {
"stivio": {
"url": "https://stivio.ai/api/mcp",
"headers": {
"Authorization": "Bearer YOUR_STIVIO_KEY"
}
}
}
}This release uses personal Bearer keys, not OAuth. Clients that only support OAuth connectors cannot authenticate directly. Model discovery and cost estimates work without a key.
Six tools, one video workflow
- list_video_models
- Compare model capabilities, durations, resolutions and credit rates. No key needed.
- estimate_video_cost
- Get the exact credit quote before creating a video. No key needed.
- get_credit_balance
- Check the credits available in your Stivio account.
- upload_image
- Upload and screen your source image or optional end frame.
- generate_video
- Generate a video from your image and motion description, or from text alone.
- get_video_status
- Follow your task and retrieve its MP4 link when it is ready.
From request to finished video
- Ask your client to list models and check your balance.
- Choose the model, duration, resolution and audio. Review the quote and approve the spend.
- For image-to-video, upload your photo; for text-to-video, provide your own prompt.
- Start generation with a unique requestId, then check the task every 10 seconds. The finished task includes the MP4 URL.
Example: “List Stivio's text-to-video models and quote a 5-second Kling O3 video at 720p without audio. Wait for my approval before generating.”
Working with images
Your client needs file access to encode and upload an image. The upload tool accepts raw base64 PNG, JPEG or WebP, up to 3 MiB decoded. A chat attachment alone may not give your client access to the original file bytes.
For clients with shell access, upload a local file using the same key, then pass the returned url to generate_video:
curl https://stivio.ai/api/upload \ -H "Authorization: Bearer YOUR_STIVIO_KEY" \ -F "file=@photo.jpg"
Keep multipart uploads under 4 MB for hosting limits. Use your client's secret handling instead of pasting a key into chat. Images are screened before generation.
Credits and results
MCP access has no separate subscription. Generation uses the same credits, content rules, watermark rules and failure refunds as the website. Identical retries with the same requestId are deduplicated for 24 hours. If a submission times out, check your video history before trying again. Never change the requestId to retry an uncertain submission: a new id can create another paid video.