YOUR AGENT. YOUR VIDEO WORKFLOW.

Connect your AI agent to Tsuyu with MCP

Bring your own compatible assistant to check videos, review timeline changes and turn an approved script into a narrated video. Keep the browser for creative controls that are not exposed to agents.

Included with paid Starter, Pro and Studio subscriptions. No separate connector fee; your existing limits still apply.

Start read-only. Add control deliberately.

Read your workflow

List your videos and configured voices, inspect render status and get result URLs. Read timeline documents where that feature is available.

Review timeline edits

Opt into edit access. Submit reviewed operations against an exact timeline version; stale versions conflict. This edits a document, not a full timeline render.

Create a text video

Opt into generation, preview the request, then approve a script with a built-in voice and static dark background. Poll the returned video for progress.

Connect in three steps

  1. Create a separate agent key

    In Settings, expand “Advanced: connect your agent”. Name the connection. Read-only access and a seven-day expiry are the defaults; enable edit or generate only when needed. Keys expire within 90 days and can be revoked in Settings.

  2. Configure your compatible MCP client

    Use Streamable HTTP with a custom Authorization header. Copy the key once into the client’s protected secret configuration, never a URL, shared prompt or repository.

    Server URL

    https://tsuyu.io/api/mcp

    Protected request header

    Authorization: Bearer YOUR_AGENT_KEY

    This is not OAuth. Clients without custom-header support are not supported yet. Agent keys cannot mint additional keys.

  3. Read first, preview before generation

    Start with list_videos or list_voices. Use preview_generation before create_video. Generation needs approved=true, max_video_allowance=1, max_ai_credits=0 and a stable idempotency key. Configure your assistant to ask you before approving; the approval flag alone does not prove human review.

Know the boundaries before you automate

An accepted text-video creation uses one video from your monthly allowance and zero AI add-on credits. Narration is included in the subscription, not free TTS or a dollar spending cap. Preview does not reserve allowance.

The current tools do not expose premium voices, AI b-roll, avatars, arbitrary uploads, full timeline rendering, social publishing, deletion or billing administration. Trial video generation stays in the browser; agents never trigger trial conversion.

Retry a completed operation with the same inputs and idempotency key to replay its result. Changed inputs conflict; an uncertain pending request is not automatically rerun. Check video status or contact support rather than changing the key to force a retry.

Agent connection FAQ

Which Tsuyu plans include MCP agent connections?

MCP agent connections are included with paid Starter, Pro and Studio subscriptions, with no separate connector fee. Your agent’s own service may charge separately. Agent requests use your existing account limits; they do not add a second video allowance. Trial accounts must create videos in the browser; agents cannot convert a trial into a paid subscription.

What can my agent create, and what does it cost?

The current generation tool creates a narrated text video with a static dark background and built-in OpenAI or local voices. An accepted text-video creation uses one video from your monthly allowance and zero AI add-on credits. Narration is included in the subscription, not free TTS or a dollar spending cap. Preview does not reserve allowance. Premium voices, AI b-roll, avatars and full timeline rendering are not available through MCP.

Can I connect any AI assistant?

Use an MCP client that supports Streamable HTTP and a protected custom Authorization header. Connect to https://tsuyu.io/api/mcp with a separately created agent key. This is not OAuth; clients that require an OAuth connection or cannot set custom headers are not supported yet.

Can an agent edit, publish or manage my account?

Read access covers your videos, render status, result URLs and configured voices. Where the timeline feature is available, optional edit access allows reviewed changes against an exact version. Generation needs a separate permission and an explicit approval flag. The flag records client intent, not proof of human review. Agents cannot publish socially, delete content, manage billing or create more agent keys.

More Tsuyu questions