Moessets

Agent-first asset catalog

Assets your coding agent can find

An agent-first asset catalog. Describe a project, get the pack that fits, pick the assets you need and fetch the files - over a remote MCP server or a plain JSON REST API that scores and explains every result.

-packs
-assets
-categories

Reading this with a fetch tool? Start at /llms.txt, which states the workflow and the live counts. In short: call GET /api/recommend?project={description}&count={n} for a coherent set, or GET /api/search?q={query}&type=pack|asset when the need is already named - both score every candidate and explain each hit. Downloading /api/packs or /api/catalog and choosing by reading descriptions works, but throws that ranking away. The counts above are filled in by JavaScript and are dashes to you; /llms.txt carries them live.

Connect an agent over MCP

The server speaks the Streamable HTTP transport of the Model Context Protocol at /mcp. The connection is stateless - no session handshake, every call stands on its own.

Claude Code, one command

claude mcp add --transport http ai-assetstore /mcp

Or as a project scoped .mcp.json

{
  "mcpServers": {
    "ai-assetstore": {
      "type": "http",
      "url": "/mcp"
    }
  }
}

Available tools

  • list_asset_packslist packs, optionally filtered
  • get_asset_packfull pack metadata plus all its assets
  • search_asset_packsfind the best pack for a query
  • search_assetsfind individual assets with relevance scores
  • get_assetone asset, optionally including the SVG content
  • get_assetsup to 20 assets in one call
  • recommend_assets_for_projectpack choice with reasoning

REST API

  • GET /api/health
  • GET /api/packs
  • GET /api/packs/{packId}
  • GET /api/assets/{assetId}
  • GET /api/search?q={query}&type=pack|asset
  • GET /api/recommend?project={description}&count={n}
  • GET /api/catalog
  • GET /files/{packId}/{slug}.{svg,png}

Packs