QuiverAI for Cursor

June 13, 2026 ยท View on GitHub

Generate, refine, and vectorize SVG assets with the hosted QuiverAI MCP server, right from Cursor.

What it does

  • Text-to-SVG generation with the Arrow family of models
  • Raster-to-SVG vectorization
  • Reference-image prompting and refinement
  • Creation and task lookup
  • A bundled agent skill that guides model selection, prompt structure, and result polling

Install

Install from the Cursor marketplace

Open the Cursor marketplace and install QuiverAI.

Manual install

If plugin-based MCP discovery is unavailable in your Cursor version, add this to .cursor/mcp.json in your project or ~/.cursor/mcp.json globally:

{
  "mcpServers": {
    "quiverai": {
      "url": "https://app.quiver.ai/mcp"
    }
  }
}

Then authenticate from Cursor's MCP settings or with:

cursor-agent mcp login quiverai

Sign in

Cursor handles authentication through its MCP login flow. The first time you call a QuiverAI tool you will be prompted to sign in to your Quiver account; tokens are stored by Cursor.

If you do not already have a Quiver account, sign up at quiver.ai.

MCP tools exposed

  • list_models - list QuiverAI models available to the connected user
  • list_creations - list generated QuiverAI creations for the connected user
  • get_creation - read integration-safe metadata for one creation
  • get_creation_content - read SVG content for one creation
  • get_task - read a generation, vectorization, or animation task for polling
  • create_generation - create a text-to-SVG generation task from text and optional image sources
  • create_vectorization - create a raster-to-SVG vectorization task from an image source
  • create_animation - create an animation task from an existing SVG creation or SVG source

Bundled skill

skills/quiverai/SKILL.md is a Cursor agent skill. It tells the agent when to call which tool, how to write effective generation prompts, how to pass direct image sources, and how to poll tasks to completion.

Example prompts

  • "Create a flat vector illustration of a delivery drone with QuiverAI."
  • "Vectorize this raster image with QuiverAI."
  • "Generate a wordmark for an indie coffee brand in a muted palette, then vectorize the best result."

Verifying the install

After installing and signing in, confirm the plugin is wired up:

cursor-agent mcp list
cursor-agent mcp list-tools quiverai

The quiverai MCP server should appear as enabled and authenticated. If your client supports listing MCP tools, it should show the eight quiverai tools above. From there, ask the agent to run the QuiverAI skill end-to-end (for example: "list models, then generate an SVG of a delivery drone, then read the SVG content").

Configuration

The plugin points Cursor at the hosted QuiverAI MCP server:

{
  "mcpServers": {
    "quiverai": {
      "url": "https://app.quiver.ai/mcp"
    }
  }
}

No further configuration is required.

Source and license

MIT-licensed. Canonical source lives in the private QuiverAI monorepo at plugins/quiverai-cursor; this repository is a public distribution artifact synced from main. Issues and feedback: support@quiver.ai.