UnrealUGC MCP

Create and manage AI UGC video ads through eleven typed MCP tools: browse available video models, creators and voices, estimate credit cost before spending, start asynchronous generations and poll for completed output. Ships as local stdio (npx -y @unrealugc/mcp) and hosted Streamable HTTP (unrealugc.com/api/mcp), MIT-licensed, with the same API key for both transports.

Server type: stdio via npm or hosted Streamable HTTP
Auth: UNREALUGC_API_KEY (usk_live_...)
Install: npx -y @unrealugc/mcp
Remote endpoint: https://unrealugc.com/api/mcp
Tools: 11 (models, creators, voices, projects, videos, credits, generation)
License: MIT
Category: Marketing / AI video ads
Built by: UnrealUGC (unrealugc.com) — registry com.unrealugc/mcp

Why This Matters for Operators

UGC-style ads convert, but shooting them is the bottleneck: finding creators, briefing, filming, editing, iterating. UnrealUGC moves the whole loop into an agent: pick a creator persona, pick a voice, estimate what the generation costs in credits before spending anything, then generate asynchronously and poll for the finished video. Eleven tools cover the entire workflow, so the agent can run "make three ad variations across three creators and tell me which performs best" as one job instead of a production calendar.

Cost-before-spend is the operator-friendly detail. estimate_video_cost prices a generation before it starts, and get_credits answers "how many 15-second videos can I make this month" — credit discipline built into the tool surface rather than discovered on the invoice.

Tools & Capabilities

Tool	What it does
get_me / get_credits	Account info and credit balance
list_models	Available video models (includes Seedance 2.0 and Kling-class models)
list_creators	Stock and custom creator personas
list_voices	Stock and custom voices
list_projects / create_project	Project organization
list_videos / get_video	Prior generations and status
estimate_video_cost	Credit estimate before generating
generate_video	Start an asynchronous generation

Tools self-describe via tools/list with full JSON Schema inputs.

Installation

Stdio (Claude Code):

claude mcp add unrealugc \
  --env UNREALUGC_API_KEY=usk_live_... \
  -- npx -y @unrealugc/mcp

Generic JSON config:

{
  "mcpServers": {
    "unrealugc": {
      "command": "npx",
      "args": ["-y", "@unrealugc/mcp"],
      "env": {
        "UNREALUGC_API_KEY": "usk_live_..."
      }
    }
  }
}

Hosted HTTP for remote MCP clients: register https://unrealugc.com/api/mcp and pass the same API key as a Bearer token.

Configuration

Variable	Required	Default	Description
UNREALUGC_API_KEY	yes	—	API key from unrealugc.com/dashboard/settings/api-keys
UNREALUGC_BASE_URL	no	https://unrealugc.com	Override for self-hosting

Business Relevance

• DTC brands produce UGC-style ad variants without a creator shoot per variation
• Growth operators A/B test creator personas and scripts at credit cost instead of agency day rates
• Agencies run client ad production as agent jobs with per-job cost estimates up front
• Lean teams cover "we need video ads" with one API key instead of a production pipeline

Integration with CorpusIQ

CorpusIQ's 40+ connectors measure what ads earn — spend, revenue, conversions across Google, Meta and storefronts. UnrealUGC produces the creative that enters those funnels. The composed loop: UnrealUGC generates and polls ad variants from an agent, the agent publishes them through the existing social stack, and CorpusIQ's ad and analytics connectors attribute results — creative production and performance measurement in one session, no spreadsheet handoff between tools.

Limitations

• New listing (repo and service launched Aug 18, 2026); no public pricing beyond credit-based billing
• Asynchronous generation means polling — long videos need the agent to check back rather than blocking
• Creator and voice libraries are stock-first; custom creators may be a higher plan tier
• Eleven tools cover the core loop, but no editing or caption tooling — post-production stays outside the MCP surface