One image and video API for your stack, plus native MCP for Claude and agents
Call text-to-image, image editing, background removal, avatars, and video from a single REST API, or wire DesignerBox straight into Claude and your agents over MCP. Pick the model per request, submit idempotent jobs with a clientRequestId, and pay per credit instead of per shoot.
Get your API key, start free
One API for every generation job
Text-to-image, editing, background removal, avatars, and video through one consistent request shape.
Generate, edit, and process images
Hit text-to-image (generate_image), image editing (edit_image), background removal (remove_background), avatars (create_avatar), and video (generate_video) through one consistent request shape. Get production access without standing up your own GPU pipeline.
Choose the model per request
Route to the right engine per call. Nano Banana Pro (14 cr) leads on high-resolution photorealism and text rendering; Seedream 5 (4 cr) and Nano Banana 2 (7 cr) trade fidelity for price, while GPT Image 2 (22 cr) and Flux (Pro 1.1 at 5 cr, Flux 2 Flex at 7 cr) round out the catalog.
MCP for AI agents and Claude
Expose DesignerBox as a tool surface your agents call directly at mcp.designerbox.ai. MCP gives LLMs a universal JSON-RPC interface to external tools, and it is supported across major agent platforms, so your agent stack stays portable.
Video generation in the same API
Generate video with generate_video through the same auth, job model, and webhooks you use for images. No separate integration required, so video lands in your product on the same plumbing as your image calls.
From API key to first asset
A simple flow from sign-up to a finished image or video URL.
Built for production integrations
Authentication and keys
Authenticate with a bearer API key scoped to your team. Rotate keys from the dashboard, and keep separate keys for staging and production so usage and limits stay clean.
Idempotent job submission
Generate a unique value client-side and send it as the clientRequestId. The server records it, so a retried or duplicated request returns the original response without re-billing, the standard pattern for safe, retryable APIs.
Webhooks and status polling
Register a webhook URL for completion events and add a periodic poll as a backstop. Track event IDs with a TTL longer than the retry window so re-deliveries are deduplicated.
Model catalog and parameters
Every request takes a model plus per-model parameters such as resolution, aspect ratio, edit masks, and reference images. Image models span Nano Banana Pro (14 cr), Nano Banana 2 (7 cr), GPT Image 2 (22 cr), Seedream 5 (4 cr), and Flux (Pro 1.1 at 5 cr, Flux 2 Flex at 7 cr), with video through generate_video.
MCP server integration
Connect the DesignerBox MCP server at mcp.designerbox.ai to Claude Desktop, Cursor, or your own agent loop. Your agent calls tools like generate_image, edit_image, remove_background, create_avatar, and generate_video natively, with the same credits and limits as the REST API.
Credits, limits, and tiers
Usage is metered in credits. FREE ($0) is for integration and testing; BASIC ($15 / 500 cr), PRO ($35 / 1000 cr), PREMIUM ($75 / 2500 cr), and ULTRA ($200 / 8000 cr) scale with throughput. Check live plan limits from the API before large batch runs.
Developer API and MCP FAQ
How do I generate images from my own application?
Sign up, get an API key, and POST a prompt with a model and a clientRequestId to the generation endpoint. For long-running jobs you get a job ID back immediately, then receive a webhook or poll the job endpoint for the finished asset URL. The same request shape covers image editing, background removal, avatars, and video.
What is MCP and why does DesignerBox support it?
MCP (Model Context Protocol) is an open standard Anthropic introduced in late 2024 that gives LLMs a universal interface to external tools. DesignerBox ships an MCP server at mcp.designerbox.ai so Claude and other agents can call image and video generation as native tools. Adoption across agent platforms has been rapid since launch.
How do idempotent jobs prevent double-billing?
You generate a unique clientRequestId and send it with each request. The server records the value, so any retry with the same clientRequestId returns the original response instead of running, and charging for, the job again. This makes network retries and at-least-once webhook delivery safe.
Which models can I call, and can I switch per request?
Yes, model selection is per request. For images you can route to Nano Banana Pro (14 cr), Nano Banana 2 (7 cr), GPT Image 2 (22 cr), Seedream 5 (4 cr), and Flux (Pro 1.1 at 5 cr, Flux 2 Flex at 7 cr); video is available through generate_video. Different models trade off quality, speed, and cost, so Nano Banana Pro leads on photorealism and text rendering while Seedream 5 comes in at a lower price.
How does pricing and credit usage work?
Usage is metered in credits across tiers: FREE $0, BASIC $15 (500 cr), PRO $35 (1000 cr), PREMIUM $75 (2500 cr), and ULTRA $200 (8000 cr). Credit cost per generation depends on the model and resolution you choose. You can query plan limits from the API before kicking off large batch runs.
Do you support webhooks or do I have to poll?
Both. Register a webhook URL to be notified the moment a job completes, and add a periodic poll as a reconciliation backstop so a missed delivery never strands a job. The recommended pattern is webhook for latency plus poll for guarantee.
Ship AI image and video into your product this week
Grab a free API key, wire up the MCP server or REST endpoint, and generate your first image in minutes. Pay per credit, switch models freely, and only scale spend when your usage does.
Get your free API key