Skip to content
MagicShot

Model Context Protocol

Make images, video and audio without leaving your assistant

Connect it once, then just ask. Your assistant makes the image or video for you, using your credits.

https://mcp.magicshot.ai/mcp

Works with

Claude Code Codex Cursor Windsurf VS Code Claude Desktop

You ask, in plain words

"Make a sunset video"

MCP

MagicShot makes it

Images Video Audio

New here? Start with this

What is MCP, without the jargon

MCP is a universal adapter for AI. Plug MagicShot in once, and your assistant gains a new skill: making images, video and voiceovers for you.

  • No code, just a one-time paste.
  • Uses your existing MagicShot credits.
  • Ask in your own words. It handles the rest.
Your AI assistant

Make me a 5 second video of a paper plane flying over a desk

Using MagicShot · text_to_video

On it. Generating a 5s clip now and I will drop the download link here when it is ready.

Rendering · 12 credits

How to use it

Connected in three steps

Around two minutes, start to finish. You only do steps 1 and 2 once.

1

Get your API key

Create one in your dashboard under API Keys.

msk_live_••••••3f9a API key
2

Paste one config

Copy the snippet for your app, drop in your key.

"url": "…/mcp"
"Authorization": "Bearer …"
3

Just ask

Type what you want. It hands back the result.

Make a sunset video

Setup

Add MagicShot to your editor

Pick your app, copy the config, and replace msk_YOUR_KEY with a key from your dashboard. Every app needs the same two things: the endpoint and your key.

Claude Code

terminal

claude mcp add --transport http magicshot https://mcp.magicshot.ai/mcp \
  --header "Authorization: Bearer msk_YOUR_KEY"

Run one command in your terminal. That is the whole setup.

Cursor

~/.cursor/mcp.json

{
  "mcpServers": {
    "magicshot": {
      "url": "https://mcp.magicshot.ai/mcp",
      "headers": { "Authorization": "Bearer msk_YOUR_KEY" }
    }
  }
}

Add to your global mcp.json, or a per-project .cursor/mcp.json.

Windsurf

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "magicshot": {
      "serverUrl": "https://mcp.magicshot.ai/mcp",
      "headers": { "Authorization": "Bearer msk_YOUR_KEY" }
    }
  }
}

Same idea as Cursor, but the key is called serverUrl.

VS Code

.vscode/mcp.json

{
  "servers": {
    "magicshot": {
      "type": "http",
      "url": "https://mcp.magicshot.ai/mcp",
      "headers": { "Authorization": "Bearer msk_YOUR_KEY" }
    }
  }
}

Note the servers key and "type": "http".

Codex

~/.codex/config.toml

[mcp_servers.magicshot]
url = "https://mcp.magicshot.ai/mcp"
http_headers = { "Authorization" = "Bearer msk_YOUR_KEY" }

Streamable HTTP. On older Codex versions, add experimental_use_rmcp_client = true at the top of the file.

Claude Desktop

claude_desktop_config.json

{
  "mcpServers": {
    "magicshot": {
      "command": "npx",
      "args": ["mcp-remote", "https://mcp.magicshot.ai/mcp",
               "--header", "Authorization: Bearer msk_YOUR_KEY"]
    }
  }
}

Bridged with mcp-remote so your Bearer key is passed through.

Any other client

the essentials

# transport  Streamable HTTP (POST + SSE)
# url        https://mcp.magicshot.ai/mcp
# auth       Authorization: Bearer msk_YOUR_KEY
# tools      suggest, list_models, text_to_image,
#            text_to_video, image_to_video, text_to_speech,
#            get_job_status, get_credits

Anything that speaks MCP and lets you set a URL and an Authorization header.

No key yet? Create a free account and grab one under API Keys.

The toolkit

What your assistant can do once connected

Generate images

Turn a prompt into finished images.

text_to_image

Generate videos

Turn a description into a clip.

text_to_video

Generate voiceovers

Turn a script into natural speech.

text_to_speech

Animate a still

Bring a photo to life with motion.

image_to_video

Pick the right model

The best model for your goal.

suggest

Browse every model

The full image, video and audio catalog.

list_models

Track a job

Follow a generation to the finish.

get_job_status

Check your balance

See credits left, anytime.

get_credits

Live demo

See it in action

One prompt in your assistant, a finished image, clip or voiceover back in seconds.

MagicShot via MCP

Try it

Things you can just ask for

Pick a prompt to preview what your assistant sends back. Images, video and voiceovers, all from plain language.

Your AI assistant

Using MagicShot ·

0:12

Ready in seconds, then downloaded or dropped straight into your project.

Model library

The latest models, all through one endpoint

Seedance 2.5

Featured

Bytedance · Video

Seedance 2.5 is ByteDance's flagship multimodal video model, available inside MagicShot on a paid plan. It turns a text prompt plus up to 50 image, video and audio references into a single clip of up to 30 seconds, with dialogue, music and sound effects generated alongside the picture.

View model

LTX 2.5

Featured

lightricks · Video

LTX 2.5 is an AI video model from Lightricks, built for speed, that turns a text prompt or a still image into a clip with synchronized audio. Inside MagicShot on a paid plan, it outputs MP4 video in portrait or landscape, up to 20 seconds long, with 4K resolution available.

View model

xAI · Image

Grok Imagine Image 2.0 is xAI's image model, launched in August 2026 as the new Quality Mode inside Grok Imagine. It builds images from text prompts and edits images you already have, at up to 2K resolution, with sharp small text and region-level changes.

View model

Flux 3

Featured

Black Forest Labs · Video

Flux 3 is Black Forest Labs' first video model, a multimodal system that generates clips up to 20 seconds long with native audio in a single pass. Inside MagicShot, it turns a text prompt, a still image, or keyframes into finished video with sound.

View model

HappyHorse 1.1

Featured

Alibaba · Video

HappyHorse 1.1 is a video generation model from Alibaba that turns text prompts and still images into short motion clips. Inside MagicShot, it produces original footage on demand so you skip the film crew, the shoot day, and the reshoot.

View model

ByteDance · Image

Seedream 5.0 Pro is ByteDance's flagship text-to-image and image-editing model, sitting above Seedream 5.0 and 5.0 Lite. It generates high-resolution images and can split a single render into 10 or more independent transparent-PNG layers you can rearrange and edit.

View model

FAQ

MCP questions, answered

MCP (Model Context Protocol) is an open standard that lets an AI assistant use outside tools. Connecting the MagicShot server simply gives your assistant new abilities: making images, video and voiceovers, checking a job, and reading your credit balance. You keep chatting like normal; it just can do more.

Any MCP client that supports remote servers with a custom header works today, including Claude Code, Codex, Cursor, Windsurf and VS Code, plus Claude Desktop through the mcp-remote bridge.

Yes. Create one in your MagicShot dashboard under API Keys and paste it into the config as a Bearer token. Keep it secret, since anyone holding the key can spend your credits. You can revoke and rotate keys anytime.

Generations made over MCP spend the same credit pool as the MagicShot app, on the plan you already have. There is no separate charge for using the server.

Not yet. Web connectors authenticate with OAuth, while the MagicShot server currently uses Bearer keys. OAuth support is planned, so editor and desktop clients are the way in for now. Claude Code is the fastest to try.

You do not have to choose. Ask your assistant to call the suggest tool with your goal and it recommends a model, or call list_models to browse everything available.

Generate without leaving your editor

Grab a free API key, paste one config, and MagicShot becomes a tool your assistant can use.

Get your API key

We use cookies to improve your experience and measure traffic. Choose what you allow.