Video Blade API

Protocols for your AI agent, storage for everything it creates, and image, video and music generation paid with tokens.

How it works

  1. Sign up at https://app.video-blade.com and top up your token balance.
  2. Pick a mode: your own local agent (Claude Code, Cursor, Codex or compatible) connected through a protocol link or over MCP, or the hosted Video Blade agent in the app (beta).
  3. Connect the agent: copy the protocol link from your dashboard (it stays on top until the agent connects, and is always in Settings) into an empty project folder and ask the agent to follow it. The agent creates the project through the API and pulls the other protocols. Claude Code, Cursor or Codex — Connect via MCP and the agent gets Video Blade as tools; other agents — the protocol link.
  4. Generate: the agent orders image, video and music generations with the model you chose for the project. Tokens are charged per actual consumption.

Authentication

Every request carries your API key from Settings, in either header:

X-API-Key: vb_…
Authorization: Bearer vb_…
curl -s https://app.video-blade.com/api/v1/me -H "X-API-Key: $VB_API_KEY"

Task lifecycle

Generations are asynchronous and belong to a record of the project: a picture series, a video session, a track, an album cover or a character photo. The order answers 202 at once with a generation id — the task was accepted, not finished. The result lands in the record itself (the series' pictures, the session's media, the track's versions), so poll the record every 15–30 seconds. Tokens ($0.005 each) are held when the task is created and settled when it completes; a failed task releases the hold. Every request, with its prompt, parameters, status, tokens and result, appears in Generation logs in the dashboard and is kept for 30 days.

The ordering calls — POST /api/v1/projects/{id}/image-series/{item_id}/generate, …/video-sessions/{item_id}/generate, …/tracks/{item_id}/generate, …/albums/{album_id}/cover/generate, …/characters/{item_id}/photos/generate — are in the API reference; each model page says where the model id goes and which parameters the model understands.

Explore