mcpField
P
Prizmad

Prizmad

Remote
prizmad.comMedia & Design1Updated 26w ago

About Prizmad

Prizmad is an MCP server for turning product URLs into customizable video ads, aimed at marketing teams and agents creating product-focused social content. It supports template selection, detailed creative controls, batch renders, and links to completed videos.

Key features

  1. 01Generate ads from product URLs
  2. 02Choose templates and avatars
  3. 03Customize captions, music, and visuals
  4. 04Render batches for variant testing
  5. 05Track progress and retrieve video links

Use cases

  1. 01Create a social ad from an Amazon or Shopify listing
  2. 02Produce multiple creative variants for testing
  3. 03Remix a previous video project

MCP support

Exposes video creation and project-management tools to MCP clients through a hosted remote server or an stdio bridge. Authentication supports OAuth flows or an API key.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://prizmad.com/api/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Packages
npm@prizmad/mcp-server2.0.3

Tools

Details

From the project README

Prizmad MCP Server

Generate AI UGC video ads from any product URL — straight from your AI agent.

npm version License: MIT Model Context Protocol Smithery

Model Context Protocol server for Prizmad. It turns any product URL — Amazon, Shopify, WooCommerce, custom store — into a platform-ready video ad in 3-8 minutes. 50+ avatars, ElevenLabs voiceover, 8 caption styles, 9 music styles, 3 CTA styles, 10 image-style presets, free-text prompt hints across image / video / music. Output is 9:16 / 1:1 / 16:9 Full HD, ready for TikTok, Instagram Reels, Facebook Ads, YouTube Shorts, Shopify, Amazon. Full commercial rights included.

This server is the agent-driver surface — pick a template, attach product images, customise the look, render, and hand back a brand-safe link. Works with Claude Desktop, Claude.ai, ChatGPT, Cursor, Zed, Continue, n8n, custom MCP SDK clients.

  • Remote MCP: https://prizmad.com/api/mcp
  • Transport: streamable-http
  • Auth: OAuth 2.1 + PKCE + Dynamic Client Registration (the "Connect" button) OR API key Bearer OR OAuth client_credentials.
  • npm: @prizmad/mcp-server

⚡ Quickstart — pick your client

Claude Desktop / Claude.ai / ChatGPT / Cursor / Zed (Connect button)

Just add a custom connector with this URL — the OAuth + DCR flow runs automatically, no install, no API key:

https://prizmad.com/api/mcp

In Claude Desktop: Settings → Connectors → Add custom connector → URL: https://prizmad.com/api/mcp. The browser opens, you sign in to Prizmad, click Authorize, the connector wires itself up.

Manual config when needed:

{
  "mcpServers": {
    "prizmad": {
      "transport": "streamable-http",
      "url": "https://prizmad.com/api/mcp"
    }
  }
}

For clients that can't drive OAuth, drop in an API key instead:

{
  "mcpServers": {
    "prizmad": {
      "transport": "streamable-http",
      "url": "https://prizmad.com/api/mcp",
      "headers": { "Authorization": "Bearer przmad_sk_live_..." }
    }
  }
}
stdio bridge (@prizmad/mcp-server) — for stdio-only clients
{
  "mcpServers": {
    "prizmad": {
      "command": "npx",
      "args": ["-y", "@prizmad/mcp-server"],
      "env": { "PRIZMAD_API_KEY": "przmad_sk_live_..." }
    }
  }
}

Or installed globally:

npm install -g @prizmad/mcp-server
{
  "mcpServers": {
    "prizmad": {
      "command": "prizmad-mcp",
      "env": { "PRIZMAD_API_KEY": "przmad_sk_live_..." }
    }
  }
}
.cursor/mcp.json snippet (project-scoped)
{
  "mcpServers": {
    "prizmad": {
      "url": "https://prizmad.com/api/mcp"
    }
  }
}

🛠 Available tools

ToolAuthWhat it does
list_templatesNoFull template catalog with features, durations, token costs.
list_avatarsNoBuilt-in avatar presets with name, gender, age, recommended voice.
recommend_templateNoTop-3 template suggestions from intent + voice / avatar / duration / budget constraints. Use this before create_video instead of guessing from the catalog.
list_my_videosYesRecent projects with projectUrl / shareUrl / downloadUrl. Find a videoId from a previous session, "remix my last video", etc.
upload_imageYesUpload an image (URL or base64) — returns a prizmad.com-hosted URL ready for productImages or avatarImageUrl.
create_videoYesStart a render. Returns videoId. Accepts caption / music / CTA / image style presets and free-text prompt hints.
get_video_statusYesSnapshot status by default; wait: true blocks server-side and emits notifications/progress until terminal (up to 10 min) — preferred over polling.
get_download_urlYesAuthenticated download URL on prizmad.com for a completed video.
create_video_batchYesLaunch 1-20 renders in parallel. Each item supports the full create_video parameter surface — perfect for A/B variant testing.

🎨 Creative customisation on create_video

Each is optional; omit any field for a randomised pick at render time.

ParamValues
captionStyleclassic, bold-impact, karaoke, pop, bounce, neon, typewriter, glow
musicStyleenergetic, friendly, professional, luxury, funny, cinematic, lo-fi, hip-hop, acoustic
ctaStyleclassic, blurred-photo, dark-solid
imageStylewarm-golden, bright-neutral, cool-diffused, window-light, earthy-ambient, studio-clean, moody-dramatic, pastel-soft, nordic-minimal, sunset-warm
imagePromptHintFree-text steer for AI creatives (≤ 400 chars).
videoPromptHintFree-text steer for AI product video clips.
musicPromptHintFree-text steer for the music generator.
language, tone, voiceId, avatarPresetId, duration, scriptStandard.

🔗 Output URLs (what to share with the user)

Every status response carries three URL kinds, in priority order:

FieldGoes to
projectUrlhttps://prizmad.com/projects/<id> — owner-only dashboard with player, remix, edit, asset library. Primary link for the signed-in user.
shareUrlhttps://prizmad.com/share/<token> — public share page. Use only when forwarding outside the account.
downloadUrlhttps://prizmad.com/api/v1/videos/<id>/download — authenticated mp4 stream proxied via prizmad.com.

The raw Vercel Blob URL is never surfaced to the agent.

🔐 Authentication

MethodWhenGet it
OAuth 2.1 Authorization Code + PKCE + DCRInteractive clients (Connect button)Automatic — no setup
API key (Bearer)Local stdio bridge, scripts, devhttps://prizmad.com/api-keys
OAuth 2.0 client_credentialsHeadless server-to-serveroauth skill

API video generation requires a Pro plan. UI generation works on any plan; the Pro gate exists only on programmatic access.

🧠 Typical agent workflow

recommend_template ─► (optional upload_image) ─► create_video
                                                  └─► get_video_status (wait: true)
                                                           └─► projectUrl + downloadUrl

Once connected you can ask your agent things like:

"Make a 30-second energetic showcase ad for this Amazon product: https://amazon.com/dp/B0EXAMPLE — moody-dramatic look, cinematic music."

The agent will pick a template, drop in the URL with the style hints, kick off the render, wait through the live progress notifications, and hand back a https://prizmad.com/projects/<id> link.

🌐 Discovery & metadata

📦 Environment variables (stdio bridge only)

VariableRequiredDescription
PRIZMAD_API_KEYYesYour Prizmad API key
PRIZMAD_BASE_URLNoAPI base URL (default https://prizmad.com)

📚 Resources

🏷 Topics / categories

mcp · mcp-server · model-context-protocol · claude · chatgpt-apps · ai-agents · oauth · remote-mcp · ai-video · video-ads · ugc-ads · ad-creative · ai-avatars · voiceover · elevenlabs · tiktok · instagram-reels · youtube-shorts · shopify · amazon · marketing

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.