Imaginode

What the imaginode MCP server does

The imaginode MCP server exposes imaginode.ai’s image and video creation capabilities through MCP. It provides access to a catalogue of 89 models, including image and video options, with pricing shown in credits and billable settings. A generation uses the credits on the connected Imaginode account and cannot exceed its available balance.

The server also exposes curated canvas templates. These cover workflows such as product imagery, renovations, video effects, children’s books, and thumbnails. Agents can inspect what a template creates, which models it uses, its approximate run cost, and the canvas link for opening the recipe.

How it works

The imaginode MCP server is hosted remotely at https://imaginode.ai/api/mcp and uses Streamable HTTP, so there is no local server package to install. An MCP client connects to that endpoint and authenticates either through OAuth 2.1 or an API key. OAuth uses authorization code flow with PKCE and supports refresh tokens; API keys are created in the Imaginode profile and sent as bearer tokens.

Image generation waits for completion and returns the resulting URL. Video generation starts a job and returns a request ID together with the exact cost. The client can then call the status tool to see whether the job is still running, completed with output URLs, or failed. Video jobs usually take one to five minutes and can take up to 30 minutes for long jobs, so polling every 30 seconds is recommended.

Setup and configuration

Create an Imaginode account with a verified email address; trial credits are included. Add the remote MCP URL in a compatible client, then complete the browser sign-in when OAuth is offered. For clients or scripts that do not support OAuth, create an API key in the profile and provide it in the Authorization: Bearer header.

The repository also includes an optional Claude Code skill. Copying skills/imaginode to ~/.claude/skills/imaginode provides usage guidance focused on cost-aware iteration, choosing cheaper models first, and polling generation jobs.

Tools and capabilities

• list_models returns image and video models, prices, and billable options.
• list_templates describes available canvas templates, their models, outputs, and approximate costs.
• get_template returns a template’s nodes, prompts, model choices, connections, example renders, and canvas URL.
• generate_image creates an image and returns its URL after completion. Its default model is Nano Banana 2.
• generate_video starts a video job and returns its request ID and exact cost. Its default model is Seedance 2.0.
• get_generation_status reports progress, output URLs, or failure with an automatic refund.

Limitations and notes

The imaginode MCP server applies a limit of 10 generations per minute to each API key and OAuth access token. OAuth access tokens last one hour and refresh silently. API keys are shown only once, can be revoked, and are limited to five active keys per account.

Input images must already be hosted in the user’s Imaginode media library or come from earlier generations. The direct REST API exposes the same model, generation, and status operations, including a public unauthenticated model catalogue. The repository is MIT licensed; the running service is hosted by imaginode.ai.