Piooy MCP server
Piooy ships a Model Context Protocol server so Claude, ChatGPT Desktop with MCP extensions, Cursor, Windsurf and other MCP-aware agents can generate images and videos inside their chat — no browser context switch.
Endpoint
- Streamable HTTP (recommended):
https://piooy.com/api/mcp/mcp - SSE (legacy clients):
https://piooy.com/api/mcp/sse - Server card: https://piooy.com/.well-known/mcp/server-card.json
Authentication
Every MCP request must carry Authorization: Bearer <piooy_api_key>. Create one at /settings/apikeys. Tokens are shown once — store them in a password manager.
Claude Desktop
Claude Desktop reads its MCP config from ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows). Add a piooy entry:
{
"mcpServers": {
"piooy": {
"url": "https://piooy.com/api/mcp/mcp",
"headers": {
"Authorization": "Bearer YOUR_PIOOY_API_KEY"
}
}
}
}Restart Claude Desktop and the Piooy tools appear in the tool picker.
Cursor / Windsurf / Zed
These clients accept remote MCP servers with bearer auth out of the box. Use the same URL and header as Claude Desktop — consult your client's documentation for the exact file location (typically mcp.json inside the settings folder).
Tools
generate_image
Start an image generation task. Returns { code, data: { taskId } }.
{
"model": "nano-banana-pro",
"prompt": "a golden retriever wearing a beret in Paris",
"resolution": "2K",
"aspect_ratio": "16:9"
}See the full model list on the image generator page or any individual facts page.
generate_video
Start a video generation task. Duration and resolution vary by model.
{
"model": "veo-3-1",
"prompt": "drone shot of a lighthouse at sunset, 16:9, cinematic",
"duration": "5",
"resolution": "1080p"
}get_task_status
Poll until the task is SUCCESS or FAILED. Image tasks typically finish in 3-10 s; video tasks in 30-120 s.
{
"taskId": "<uuid returned by generate_image or generate_video>"
}Credits and billing
Credits are deducted from the account that owns the API key. The same quotas and rate limits that apply to /api/v1/generate apply here. Unlimited-plan subscribers can call free-tier models at zero cost.
Raw REST alternative
If your agent framework does not speak MCP yet, call the same logic over plain HTTPS. See the OpenAPI spec for complete schemas.