FluxNote MCP server for AI video.

1. Check access and create a key

Open Developers in the app to check MCP availability and create a scoped key. Only tools allowed by the key’s scopes are advertised. Access must be enabled for your account.

2. Connect with Streamable HTTP

Use the endpoint displayed in the app. The hosted endpoint is https://api.fluxnote.io/mcp when enabled. Configure Authorization as a private Bearer header. This example uses the mcpServers format; settings vary by client.

{
  "mcpServers": {
    "fluxnote": {
      "type": "http",
      "url": "https://api.fluxnote.io/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Requires a client supporting Streamable HTTP and custom authentication headers. OAuth-only connectors and stdio are not supported. Do not place real keys in chat, URLs or committed configuration.

3. Estimate, create, then poll

1. Use list_voices and get_options to choose valid settings.
2. Call estimate_video and approve the cost before creating.
3. Call create_video with exactly one of prompt or script and an idempotency_key.
4. Poll get_video until completed or failed. Download the returned media_url.

Tool calls with sufficient permissions execute immediately. Ask your client to require confirmation for writes; confirmation is not enforced by the server.

Images, clips, music and more

• create_image and upscale_image: images from a prompt or reference images, and 2× or 4× upscales.
• create_clip: a short AI video from a prompt or a start image.
• create_music (Rise and above): an original track from a prompt, with or without lyrics.
• create_voiceover: a script narrated in a voice from list_voices; returns the MP3.
• create_lipsync (Rise and above): make a face image talk, or re-sync a video, to audio imported with import_asset.
• create_audiobook: chapters narrated into an audiobook within your plan’s monthly allowance. Poll get_audiobook.

Image, upscale, clip, music and lipsync tools return an ID: poll get_generation until it completes. Prices are listed by get_options under image_models, upscale_models, clip_models, music_models and lipsync_models. These tools need a key with the Studios or Audiobooks permissions.

Tools and results

38 tools cover account and credits, voices and options, videos, images, AI clips, music, voiceovers, lipsync, audiobooks, assets and operations, Autopilot series, connected accounts, posts, calendar and analytics. Arguments mirror the REST API, with path and query fields alongside body fields.

Results include http_status, request_id and result. Application errors set isError. Imports and caption jobs return operation IDs: poll get_operation. A completed operation with response_status ≥ 400 failed.

Retries and safety

Respect retry_after_seconds and use the same idempotency_key with the identical payload when recovering a request. Never resubmit an uncertain write under a new key without checking its operation or resource. Renew expired asset URLs using get_asset. Publishing and active series can spend credits and change real content.