SpicyAPI MCP Server

official

Official SpicyAPI MCP server (github.com/spicyapi-ai/spicy-mcp). Let Claude, Cursor or Codex generate images, videos, speech and music with 200+ models through an unrestricted AI API with no platform content filter, with an exact USD quote and your confirmation before every paid step.

View on spicyapi.aiTry this MCPSponsored

What can you do with SpicyAPI MCP?

  • Generate uncensored media — Ask for text-to-image, text-to-video, or image-to-video creations using frontier models without content filters or approval queues.

  • Edit and transform visuals — Request face swaps, image editing, or turn a still image into a short video clip with sound, up to 30 seconds.

  • Run chat completions — Use OpenAI-format chat with streaming support, switching between models by changing just the model name.

  • Manage media jobs — Submit generation tasks via createTask, then poll for results or receive webhook callbacks when complete.

  • Control data retention — Set custom expiration times for generated files, prompts, and uploaded references, or destroy task outputs outright.

Documentation

Frontier models, uncensored editions included.

Uncensored

Uncensored edition: Wan 3.0 Spicy

Turn text, a still or ten references into 2 to 30 seconds with sound.

Open the model

1 / 6

More than an API.

A studio for creators, one API for developers, tools for AI agents and rankings for anyone choosing a model — all on one account.

For creatorsSpicy Studio

Make images and short videos right in the browser — no code, no platform filter. Start from a style or an effect, then keep shaping it on the canvas.

Open Studio

Developer API

Agents & MCP

Rankings

Spicy & LoRA

Nobody askswhat you are making

Pick by task, read what the model actually accepts, and see the price before you send anything. No declaration, no queue.

Browse every model

Why developers bring the unfiltered work here

Uncensored, and upfront about it

We run no content classifier, ask for no per-request declaration and keep no approval queue. Spicy editions are built for unrestricted generation, and each model page states what that model accepts.

Browse uncensored models

One API for video, image and chat

Media jobs use createTask with polling or callbacks; chat speaks the OpenAI format, streaming included. Switching models means changing one name, not your integration.

Read the quickstart

Pay per call, settled on actual use

Each model bills in its own unit — per image, per second or per million tokens. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

See pricing

You decide how long any of it stays

Three clocks you set yourself: generated files, the prompt text behind them, the references you upload. Or destroy a finished task's outputs outright — the charge stays on record, the content does not.

What we keep, and for how long

Nothing here is locked down.

Not the models, not the way you call them, not the billing, and not what you leave behind. Four places a platform usually keeps a hold on you; four places we do not.

  1. The models are not filtered. No content classifier, no per-request declaration, no approval queue. Each model page states what that model itself accepts.
  2. The integration is not locked in. Native REST, an OpenAI-compatible surface, our own SDK, CLI, MCP server and Agent Skill — or whatever client you already use, pointed at a new base URL.
  3. The spending is not committed. No plan, no pack, no subscription. One key and one balance you top up: each run takes its listed price, and you can cap the key per day, per month and for its lifetime.
  4. The traces are not permanent. Your own clock on generated files, prompt text and uploaded references — never past our published maximum, as short as an hour. A finished task's outputs you can destroy outright, whenever you say.

Charged in each model's own unit — per image, per output second, per million tokens. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting.

Pricing

Ways in

One key works across all of them. Bring the client you already use, or take one of ours.

Underneath it is one HTTP API: createTask, then poll or let a webhook tell you. Chat also speaks the OpenAI format, so an existing client only needs a new base URL.

Developer hub

Three steps to the director's chair

An account, a balance, a key. Between this page and your first shot that nobody vetted, there is nothing else.

  1. Create your account

    One account covers the whole catalogue. Sign up with an email address and a password, or carry on with an identity you already have. Create an account

  2. Top up by card or crypto

    Top up by card or in crypto, USDT included. Failed or timed-out tasks are usually not charged. Some models charge for content refusals as stated on their model page; check those billing terms before submitting. See what each model costs

  3. Create an API key

    Generate a server-side key in the console and give it a daily spend cap. Then send the same JSON you tried in the Playground, with that key in the Authorization header. Read the quickstart

Straight answers

Not covered here? Ask us directly — technical, billing or content questions all reach a person.

Contact us

Go straight to a model family

The families we lead with for video, image and text, in catalogue order. Each one opens on its endpoints, its input schema and its current price.

View all models