Moda raises $7.5M from General Catalyst to build AI design agent with fully editable canvas — not 'AI slop'

Summary

Moda, a five-person startup, raised $7.5M from General Catalyst to build an AI design agent on a fully editable vector canvas. This is a deliberate technical departure from every other AI design tool on the market.

Founder Anvisha Pai argues that most AI design tools produce what she calls "slop" — templated outputs trapped inside rigid drag-and-drop interfaces that users can't meaningfully edit. Moda's bet is that the right AI design tool isn't one that generates a million ad creatives in thirty seconds, but one that lets users create things they couldn't have built otherwise. That ambition requires a real vector canvas where elements can be placed precisely, moved freely, and drawn from scratch, not an HTML output layer or image generator.

Getting there took months and five failed approaches before the underlying agent worked. The agent writes directly to a vector canvas rather than generating image outputs or HTML, giving users pixel-level control over the output.

Early traction

Pai describes inbound demand as "insane," with early users ranging from investment banking analysts to a sheepskin farmer in New Zealand. Support volume has been high enough that she's been dropping Google Meet links into support threads and jumping on calls in real time.

A more telling signal: a user at a UK firm independently connected Moda to Claude via an undocumented MCP integration, embedding the entire workflow into Claude skill files and chaining the tools together without any official documentation. Pai had no idea how the user figured it out.

That pattern is pushing Moda to treat its API as a first-party product from day one. She says she would never have done this early in a consumer product under normal circumstances. The demand for chaining, automation, and MCP connectivity is unlike anything she's seen in prior product cycles.

Competitive pressure

Google launched Stitch last week, a tool that transforms ideas into UI designs for mobile and web. At five people, Moda can't out-resource Google. Pai's answer is that the vector canvas architecture and the quality ceiling it enables are a different category than what Stitch or generic image-to-design tools are attempting. She draws an analogy to Claude Code: tools that expand capability rather than just compress time.