Kie.ai MCP Server

MCP server for the Kie.ai API — unified access to 25+ AI models for image, video, and audio generation.

Models

Image

• Nano Banana 2 (Google Gemini 3.1 Flash) — photorealistic, 4K, text rendering
• Ideogram V3 — text-to-image, edit, remix, reframe (flat design, typography, logos)
• Flux Kontext Pro/Max — consistent style, editing
• Flux 2 Pro/Flex — multi-reference, text rendering
• Seedream V4/V5 Lite — detail fidelity, text
• GPT-4o Image — versatile generation and variants
• Qwen — bilingual prompts
• Z-Image — ultra-fast, bilingual text
• Grok Imagine — image + video
• Midjourney v7/v6/niji6 — artistic, stylized
• Topaz — AI upscaling 1x-8x
• Recraft — background removal

Video

• Kling 2.6/3.0 — text/image to video, avatars
• Seedance (ByteDance) — lite/pro quality
• Hailuo — standard/pro
• Sora 2 — OpenAI video
• Veo3 (Google) — professional quality
• Wan — text/image/video to video, animate
• Runway Aleph — video generation
• InfiniTalk — lip-sync talking avatars

Audio

• Suno — AI music generation
• ElevenLabs — text-to-speech, sound effects

Setup

npm install
npm run build

Claude Code

claude mcp add kie-ai -- node /path/to/kie-ai-mcp/dist/index.js

Or add to ~/.claude/settings.json:

{
  "mcpServers": {
    "kie-ai": {
      "command": "node",
      "args": ["/path/to/kie-ai-mcp/dist/index.js"],
      "env": {
        "KIE_AI_API_KEY": "your-api-key"
      }
    }
  }
}

Environment Variables

Variable	Required	Description
KIE_AI_API_KEY	Yes	API key from kie.ai/api-key
KIE_AI_CALLBACK_URL	No	Default webhook URL for task completion

Usage

All generation tools are async — they return a taskId immediately. Use get_task_status to poll for results.

1. Call a generation tool (e.g., ideogram_v3_generate)
2. Get back a taskId
3. Poll with get_task_status until state = "success"
4. Result URLs are in the response (expire after 14 days)

Tools

Tool	Type	Description
nano_banana_image	Image	Gemini 3.1 Flash generation/editing
ideogram_v3_generate	Image	Ideogram V3 text-to-image
ideogram_v3_edit	Image	Ideogram V3 mask-based editing
ideogram_v3_remix	Image	Ideogram V3 image remixing
ideogram_v3_reframe	Image	Ideogram V3 aspect ratio adaptation
flux_kontext_image	Image	Flux Kontext Pro/Max
flux2_image	Image	Flux 2 Pro/Flex
openai_4o_image	Image	GPT-4o image generation
qwen_image	Image	Qwen image generation
z_image	Image	Z-Image ultra-fast
grok_imagine	Image	Grok Imagine (image + video)
midjourney_generate	Image	Midjourney v7/v6/niji6
topaz_upscale_image	Image	Topaz AI upscaling
recraft_remove_background	Image	Background removal
kling_video	Video	Kling 2.6/3.0
kling_avatar	Video	Kling AI avatars
bytedance_seedance_video	Video	Seedance lite/pro
hailuo_video	Video	Hailuo standard/pro
sora_video	Video	Sora 2 / Sora 2 Pro
veo3_generate_video	Video	Google Veo3
wan_image	Image	Wan 2.7 Image
wan_image_pro	Image	Wan 2.7 Image Pro
wan_video	Video	Wan text/image/video
wan_animate	Video	Wan animate images
runway_aleph_video	Video	Runway Aleph
infinitalk_lip_sync	Video	InfiniTalk lip sync
suno_generate_music	Audio	Suno AI music
elevenlabs_tts	Audio	ElevenLabs TTS
elevenlabs_ttsfx	Audio	ElevenLabs sound FX
get_task_status	Utility	Check task progress
get_credit_balance	Utility	Check credit balance

Rate Limits

• 20 requests per 10 seconds per account
• Supports 100+ concurrent tasks

License

MIT