Animx for Claude

1. In Claude, open Settings > Connectors > Add custom connector and paste the connector URL above.
2. Claude starts an OAuth sign-in. Use your Animx account — the tools read and create your generations only.
3. Ask for an image or video in chat. The result renders inline on the Animx media card.

No API key or developer setup is needed — free and paid Claude accounts can both add it. Works in claude.ai chat, Claude Desktop, mobile, Cowork, and Claude Code. When Claude calls a tool for you, credits are deducted from your Animx account the same as in the Animx app. Your plan's usual limits apply.

• generate_image — text-to-image. Returns a generationId immediately; the media card updates itself when the render finishes, usually in under a minute.

  • prompt — required — describe the scene, subject, motion and camera style.
  • model — optional — e.g. 'Nano Banana 2', 'Seedream', 'Grok Imagine'. Defaults to Nano Banana 2.
  • aspectRatio — optional — 'W:H' such as '1:1', '16:9', '9:16'. Defaults to '1:1'.
  • resolution — optional — '1K', '2K' or '4K' where the model supports it.
  • imageUrl — optional — public HTTPS URL of an image to edit or extend (image-to-image).
  • referenceImages — optional — up to 4 reference images (public HTTPS URLs) to condition the result on: png, jpeg, webp or gif.

• generate_video — text-to-video and turning a still image into motion. Videos render asynchronously, usually taking 30–90 seconds, longer when the queue is busy.

  • prompt — required — the scene, subject, motion, and camera style.
  • model — optional — defaults to 'Seedance 2.0 Mini'. Only some models render audio; the tool lists the currently available models.
  • aspectRatio — optional — 'W:H'. Defaults to '16:9'.
  • duration — optional — seconds between 2 and 30. Defaults to 8; longer clips cost more credits.
  • resolution — optional — '480p' to '4K', model-dependent.
  • imageUrl — optional — animates this image as the first frame (image-to-video).
  • referenceImages — optional — up to 4 public HTTPS URLs; only supported by models with reference endpoints.

• get_generation — status, progress, and the finished media URL. Gives the most recent generation when called without an id, so "check my last generation" always works.

  • generationId — optional — omit it to check the most recent generation.

  When status is done, the response carries a public, runnable outputUrl. For pending or processing generations a queue field reports the job's position and whether it is actively rendering — a 0% progress with queue information means the generation is waiting or rendering normally, not stuck. Generations that were never picked up by a worker are marked errored after 25 minutes and their credits are refunded automatically.

Every generation tool returns an inline card that shows the finished image or video with Open and Copy link buttons, plus the direct URL. The card updates on its own while Claude renders, so you normally never need to ask for status — just wait for the card. Ask out loud only when you want a status check or a fresh download link; Claude then calls get_generation for you.

Follows your light/dark mode: the card picks up the conversation's theme both on claude.ai and on desktop.

When you generate, Animx receives your prompt, any image URLs you point it at, and your Animx sign-in, delivered over HTTPS to animx.ai. It stores the result in your account under your existing Animx terms, reads back only your own generations, and shares nothing with third parties beyond model providers. No API keys are stored anywhere by this connector.

Full details live in the privacy policy and terms of service. Questions or problems: contact Animx.

The connector authenticates with plain OAuth 2.0 + PKCE using dynamic client registration, uses no custom request headers, and requires no third-party services beyond animx.ai itself.