Overview
Put Imagifly image and video generation inside your app, your scripts and your agents.
Imagifly is an AI image and video generation platform. One credit wallet covers every model: turn a prompt, a reference image, or a first frame into cinematic stills and short videos, then keep the results in the same library you use in the app.
This site is the integration guide for that platform. You can call it over HTTP, or connect it to an agent over MCP and let the agent call it for you.
What you can build
Image generation
Text-to-image with optional reference images, aspect ratios, quality tiers and reasoning steps.
Video generation
Text-to-video, reference-to-video and first/last-frame workflows across the video models.
Agent workflows
One MCP server that exposes the same generation tools to Claude Code, Codex, OpenClaw, WorkBuddy and Cursor.
How it works
- Create an API key in the API console. Treat it like a password.
- Send a generation request over HTTP, or add the MCP server to your agent and let it call the tools.
- The request returns immediately with a task ID; poll that ID until the task finishes.
- Read the output URLs from the finished task and download the files.
Async by design
Generation is asynchronous. A 202 response means the task was accepted and credits were reserved, not that the image or video exists yet. Query the task until status becomes success or failed.
Start here
API conventions
Create a key, send your first request and read the rules every endpoint follows.
Connect an agent
Install the Imagifly MCP server in Claude Code, Codex, OpenClaw, WorkBuddy or Cursor.
Errors
Status codes, error identifiers and which failures refund credits.
What the API covers
- Generation — submit tasks, poll status, read output URLs.
- Models — the live model catalog, including per-model credit prices, supported sizes, aspect ratios and reference limits.
- Account — the API key identifies the account; every task counts against that account's credits.
There is no endpoint for listing, deleting or cancelling tasks, and no webhook or callback. Polling is the supported way to wait for a result.