HiAPI

What is HiAPI?

HiAPI is a unified API gateway for AI image, video, and audio generation. Developers and AI agents call leading models like GPT Image 2, FLUX, Seedance, and MiniMax through one endpoint, with persistent artifact links and optional callbacks. It handles long-running tasks, cost-aware routing, and supports MCP and Skills for agent integration. Ideal for teams building media generation tools or creative automation pipelines.

What are the features of HiAPI?

• Unified Task API: One POST endpoint (/v1/tasks) for all models. Pass model name and input; get back a task ID for polling or callback.
• Persistent Storage: Set storage: "persistent" to receive durable artifact links instead of temporary URLs. No extra storage infrastructure needed.
• AI Agent Ready: Supports MCP tools, Skills, and llms.txt so AI agents (Claude, Cursor, Codex) can discover and call HiAPI workflows.
• Availability-First Routing: Choose routes like default, beta, or ext to balance cost and availability. Designed for stable access, not just cheapest price.
• Callbacks & Idempotency: Submit tasks with a callback URL and an idempotency key. Prevents duplicate tasks and avoids polling for completion.
• Transparent Per-Task Pricing: Each model lists its price upfront (e.g., GPT Image 2 from $0.007/image). Volume discounts available for larger usage.

What are the use cases of HiAPI?

• AI video SaaS teams testing multiple video models and keeping generated assets reachable without building storage per provider.
• Creative automation teams passing persistent artifact links between campaign tools, review, and downstream automation.
• Agent workflow teams requesting, waiting, retrieving, and reusing generated media through a clean task API with callbacks.
• Developers integrating image generation (e.g., product mockups, social posts) with a single API key and predictable cost.

How to use HiAPI?

• Sign up at hiapi.ai to get a free API key with up to 2,000 credits.
• Call POST /v1/tasks with your API key, model name (e.g., gpt-image-2/text-to-image), and input parameters.
• Optionally set storage: "persistent" to get durable artifact links, and provide a callback URL to be notified on completion.
• Poll GET /v1/tasks/:id to check status, or wait for the callback. Retrieve the generated media from the returned links.
• Test models in the Playground on each model page before writing code.