Build with AI

For AI agents

The Vivideo API is designed to be called by LLMs and autonomous agents, not just humans. Consistent JSON, stable error codes, and machine-readable specs make code generation reliable.

Machine-readable entry points

i
One spec, always in sync. The interactive reference, the OpenAPI file and the text summary are all generated from the live implementation — so what an agent reads is what the API actually does.

Guidance for agents

If you're building an agent that generates videos on a user's behalf, follow these rules:

  • Check GET /v1/me first — read plan, premium and credits before an expensive call, and surface the returned links (upgrade / buy credits) to the user when relevant.
  • Always send an Idempotency-Key on POST /v1/videos so a retried call never creates or charges twice. On 409 duplicate_request, reuse details.existing_video_id.
  • Poll at poll_interval_seconds (usually 5s), not in a tight loop. Rendering takes seconds to a few minutes — set the user's expectations.
  • Branch on error.code, never on message text. Every error tells you whether it's retryable.
  • Handle the free-tier preview — a status: "preview" response means the account is free; show preview.image_url (blurred) and the upgrade_url, then render with POST /v1/videos/{id}/render after upgrade.

Minimal agent loop

pseudocode
account = GET /v1/me
if not account.premium and not account.free_video_available:
    tell user → account.links.upgrade

video = POST /v1/videos  { mode, model, prompt, ... }   # + Idempotency-Key
while video.status in ("queued", "processing"):
    wait video.poll_interval_seconds
    video = GET /v1/videos/{video.id}

if video.status == "completed":  deliver video.video_url
elif video.status == "preview":  show preview.image_url + upgrade_url
else:                            report video.error.message  # credits refunded

MCP, CLI & Skills

This REST API is the foundation. A Vivideo MCP server, CLI, and agent Skills are built on top of it — so the same capabilities are available natively inside Claude, Cursor, terminals and your own agents. More here soon.


Machine-readable: interactive reference · openapi.yaml · llms-api.txt. Base URL https://api.vivideo.ai.