Amplibee
What is SpicyAPI?
SpicyAPI is a pay-as-you-go API for image, video, audio and chat models, with a browser studio for people who prefer not to write code.
Video, image and OpenAI-compatible chat models share one API key. Media generation uses a common task lifecycle with asynchronous jobs, polling and webhooks, and chat supports streaming in the OpenAI format.
The studio lets you make images and short videos in the browser, with a canvas, a video editor, effects, styles and a prompt library.
What can you build with it?
Anything that needs generated media or chat behind one key.
- Text to image, image editing, text to video, image to video and face swap, picked by task
- Chat through an OpenAI-compatible endpoint with streaming
- Audio models, including speech and music
- Your own LoRA on supported video and image models
- Rankings and a blind arena for choosing a model by what it actually makes
How does it work with coding agents?
SpicyAPI publishes an MCP server, a CLI, SDKs and an agent skill, all under the MIT license.
The MCP server lets assistants such as Claude, Cursor, VS Code and Codex browse models, compare prices and run jobs. The docs say it adds 22 tools and that every paid step waits for your confirmation.
Public SDKs cover TypeScript, Python, Go, PHP and Java. There are also ComfyUI custom nodes and an n8n community node. These client projects are separate from the hosted service.
What does it say about content limits?
SpicyAPI says it runs no content classifier and keeps no approval queue.
Its home page says spicy editions are built for unrestricted generation, and that each model page states what that model accepts. The pricing page says SpicyAPI forwards requests without adding a platform content scan, and that legal and consent duties remain with the customer.
If your product serves the public, check each model page and your own legal obligations before you build on it.
What data does it keep?
You set how long generated files, prompt text and uploaded references are kept.
The home page describes three retention clocks you control, and says you can destroy a finished task’s outputs outright. The charge stays on record, the content does not.
How much does it cost?
You top up a US-dollar balance and pay per call, with no subscription.
The pricing page lists each model family once with its lowest published price, such as from $0.019 per image for one editing model. It says failures are usually not charged, that you can top up by card or crypto, and that the public clients are free while hosted inference is paid. Open a model page to price your own settings before you run.
Alternatives to SpicyAPI
Several services host image, video and chat models behind one API. SpicyAPI combines media and chat models, a studio and agent tooling, and states plainly that it adds no content filter.
- falA hosted platform for running image, video and audio models through an API.fal.ai ↗
- ReplicateA platform for running open and community models, including image and video, through an API.replicate.com ↗
- OpenRouterA single API for many language models from different providers.openrouter.ai ↗
How SpicyAPI compares
| Option | Best for | Trade-off |
|---|---|---|
| SpicyAPI | Developers and creators who want image, video, audio and chat models behind one USD balance, with agent tooling | A new service from October 2026, and content responsibility sits with you, so read each model page |
| fal | Running generative media models in production through one API | Focused on media, so check its current model list and content policy |
| Replicate | Trying community models and publishing your own | Pricing varies by model and hardware, so estimate your cost per run |
| OpenRouter | Switching between chat models with one key | Centered on text models rather than image and video generation |
SpicyAPI may suit you if you want media and chat models behind one prepaid balance and an MCP server for your coding agent. Because it is new and does not filter content itself, read the model page, the pricing and the data retention terms before you rely on it.
Questions people ask about SpicyAPI
What is SpicyAPI?
A pay-as-you-go API and studio for image, video, audio and chat models, with one key for the whole catalog.
How many models does it offer?
The home page counts 202 models: 76 video, 83 image, 26 text and 17 audio.
Is SpicyAPI free?
No. Hosted inference is billed by usage from a prepaid USD balance. The SDKs, CLI and MCP client are free MIT-licensed projects.
Is there a subscription?
No. The site says you pay per call in US dollars, with no subscription or credit conversion.
Is the chat API compatible with OpenAI?
Yes. Chat uses the OpenAI format with streaming, so existing clients can be pointed at it.
Does it have an MCP server?
Yes. @spicyapi/mcp works with Claude, Cursor, VS Code, Codex, Windsurf and Gemini CLI, and asks you to confirm every paid step.
Which languages are the SDKs in?
TypeScript, Python, Go, PHP and Java, plus a CLI.
Does it filter content?
SpicyAPI says it adds no platform content scan and that each model page states what that model accepts.
Can I use it without writing code?
Yes. Spicy Studio makes images and short videos in the browser.
What are the alternatives to SpicyAPI?
fal, Replicate and OpenRouter all offer hosted model APIs.
