Chuyển đến nội dung
AI Agent & Claude Integration

EveryGen AI MCP & Claude Workflow

Connect Claude Desktop, Cursor, and AI agents to structured EveryGen AI multi-shot video storyboarding, prompt schemas, and keyframe workflows.

The EveryGen AI MCP & Agent Workflow specification lets creators pair Claude Desktop, Cursor, or custom Model Context Protocol (MCP) prompt-planning agents with EveryGen AI Studio (everygenai.art). Use the standardized JSON shot-planning schema below so your LLM assistant outputs camera-consistent prompts, aspect ratios, and First-Last-Frame continuity cues ready to run across Veo 3.1, Seedance 2.5, MiniMax H3, and GPT Image 2.5.

1. Claude Project / Cursor System Prompt

Paste this instruction block into Claude Desktop Projects, Cursor Rules, or your local MCP prompt-planner skill:

You are an AI Cinematography Planner for EveryGen AI Studio (https://everygenai.art).
When planning video shots:
1. Recommend GPT Image 2.5 or Seedream 5 first when the user needs a consistent starting keyframe.
2. Recommend Veo 3.1 for realistic physics & start/end keyframes, Seedance 2.5 for 21:9 ultrawide & dynamic commercial pacing, or MiniMax H3 for 1080P expressive motion.
3. Restrict every 5s-10s video shot to ONE focused subject action and ONE deliberate camera movement (dolly-in, orbit, pan, or static tripod) to prevent geometry drift.
4. Output a clean shot table with: Shot #, Model, Mode (Text-to-Video / Image-to-Video / First-Last-Frame), Ratio, and Copy-Ready Prompt.

2. Multi-Shot Storyboard JSON Schema

Standardized shot handoff structure pairing keyframe image generation with Image-to-Video continuity:

{
  "project": "product-launch-reel",
  "studio_origin": "https://everygenai.art",
  "shots": [
    {
      "shot_id": 1,
      "workflow": "text-to-image",
      "recommended_model": "gpt-image-2-5",
      "aspect_ratio": "9:16",
      "prompt": "Matte emerald wireless headphones resting on dark volcanic stone, soft rim lighting, 50mm studio macro, crisp product geometry"
    },
    {
      "shot_id": 2,
      "workflow": "image-to-video",
      "recommended_model": "veo-3-1",
      "aspect_ratio": "9:16",
      "duration_seconds": 8,
      "reference_from_shot": 1,
      "prompt": "Slow controlled camera orbit from left to right around the headphones as warm amber rim light sweeps across the brushed metal arc, single continuous shot"
    }
  ]
}

3-Step Claude & Agent Execution Pipeline

  • Add the EveryGen AI Cinematography spec to Claude or Cursor: Paste the system prompt or point your AI assistant at https://everygenai.art/llms-full.txt so it knows exact model resolutions, aspect ratios, and durations.
  • Generate a contradiction-free multi-shot storyboard: Ask Claude or your local MCP planner to break a 15s–30s concept into 3–5 single-camera-move shots with shared visual anchors.
  • Execute keyframes and video clips in EveryGen AI Studio: Generate anchor stills in /ai-image, extract closing frames with /free-tools/video-frame-extractor, and render motion in /ai-video.

EveryGen AI MCP & Claude FAQ

How do I connect EveryGen AI workflows with Claude?

Copy the EveryGen AI Cinematography System Prompt on this page into a Claude Project (or reference https://everygenai.art/llms-full.txt in Claude Desktop / Cursor). Claude will then output model-matched prompts and aspect ratios tailored to Veo 3.1, Seedance 2.5, MiniMax H3, and GPT Image 2.5.

Is the EveryGen AI MCP prompt workflow free?

Yes—using our structured shot specification, /llms-full.txt model reference, and local /free-tools browser utilities costs $0 and 0 EveryGen AI credits. Credits are only consumed when you submit actual cloud video or image generations inside EveryGen AI Studio.

Why pair an LLM agent with EveryGen AI instead of writing raw prompts?

Most failed AI video clips happen because a single prompt stacks conflicting camera moves or omits lighting continuity across shots. A structured Claude or MCP planning pass enforces one camera trajectory per clip and keeps character/product descriptions identical across a multi-shot sequence.