QuiverAI

Use this skill for QuiverAI API integration work. Start with direct API calls or the official Node.js SDK. Escalate to a remote MCP server only when the job needs reusable live tools, centralized auth, or a shareable agent surface.

Workflow

1. Read references/api.md for the current API surface and official doc links.
2. If the user wants a hosted agent surface, read references/remote-mcp.md and use $building-mcp-server-on-cloudflare.
3. Choose the smallest viable integration surface:
   • Node.js project: prefer @quiverai/sdk.
   • Other languages or thin integrations: call the REST API directly.
4. Load auth from QUIVERAI_API_KEY. Keep the key in an environment variable or secret manager. Never commit it.
5. Start with the smallest useful endpoint:
   • GET /v1/models to discover available model IDs
   • POST /v1/svgs/generations for text-to-SVG
   • POST /v1/svgs/vectorizations for image-to-SVG
6. Default to non-streaming calls. Use stream: true only when the caller explicitly needs server-sent events.
7. Handle failures deliberately:
   • respect Retry-After and X-RateLimit-*