The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Imaginode listing page.
Generate AI images and videos from ChatGPT, Claude, Claude Code, Cursor, or any MCP client, powered by imaginode.ai, the node-canvas AI creation studio.
https://imaginode.ai/api/mcp (Streamable HTTP).The API returns the exact cost of every generation, but you often want the number before writing the call. Two public pages answer that without an account:
GET /api/models: the same catalogue and
prices as machine-readable JSON, no auth, so a client can build its own
estimate.ChatGPT: Settings, Apps and connectors, Create: paste https://imaginode.ai/api/mcp, choose OAuth, sign in to your Imaginode account.
Claude.ai / Claude Desktop: Settings, Connectors, Add custom connector: paste https://imaginode.ai/api/mcp and sign in when Claude asks.
Claude Code (the sign-in opens in your browser on first use):
Cursor (.cursor/mcp.json, OAuth):
With an API key instead (scripts, clients without OAuth): create a key in your profile, section API keys, and send it as a header:
Then just ask your assistant: "Generate a cinematic image of a red fox in a misty forest", "Make a 5-second video of ocean waves at sunset", or "Which Imaginode template turns my selfie into a collectible figurine?".
| Tool | What it does |
|---|---|
list_models | Catalogue of image and video models with prices in credits and billable options. |
list_templates | The curated canvas templates: what each produces, its models, the approximate cost of one run. |
get_template | Full recipe of a template (nodes, models, prompts, wiring, example renders) and the link that opens it in the canvas. |
generate_image | Generates an image and waits for the result: the URL is returned directly. Default model: Nano Banana 2. |
generate_video | Starts a video generation and returns a request_id plus the exact cost. Default model: Seedance 2.0. |
get_generation_status | Follows a generation (running, completed with output URLs, or failed with automatic refund). |
Videos typically take 1 to 5 minutes (up to 30 for long jobs): poll get_generation_status every 30 seconds.
https://imaginode.ai/.well-known/oauth-authorization-server and https://imaginode.ai/.well-known/oauth-protected-resource.generate. An unauthenticated request gets a 401 with WWW-Authenticate: Bearer resource_metadata="…".The MCP server is a thin facade over three public endpoints you can also call from any script (API key or OAuth access token in the Authorization header):
GET https://imaginode.ai/api/models : model catalogue and prices, no auth.POST https://imaginode.ai/api/generate : body { "model", "prompt", "duration", "resolution", "aspectRatio", "audio", "imageUrls" }, header Authorization: Bearer imk_…. Returns { requestId, cost, balance }.GET https://imaginode.ai/api/generate/status?requestId=… : returns running, completed (with outputUrls) or failed (refunded).Input images must be media hosted on your Imaginode account (media library or previous generations).
Full documentation: imaginode.ai/docs/api-mcp.
The skills/imaginode folder contains an optional Claude Code skill that teaches Claude how to use this server well (cheap-model-first iteration, cost awareness, polling cadence). Install it by copying the folder to ~/.claude/skills/imaginode.
MIT. The server itself runs at imaginode.ai; this repository hosts its public manifest, documentation and skill.