How to Win the Next Decade of Design Tooling

The designers building some of the most ambitious products right now are reaching for Figma less and Claude Code, Codex, and Cursor more. 

I've seen this pattern emerge across conversations over the past few months, and it was on full display at the Design Night we recently co-hosted with Vercel. Cynthia, a designer at Block, runs Claude Code with a Figma MCP as part of her daily workflow. Abigail at Wispr now spends more than half her week inside GitHub branches via Dessn. Chris at Harvey uses Cursor to spin up new product iterations. Pablo at Vercel uses v0 to prototype. And Paper (one of the most talked-about new design tools in the ecosystem) is something most of the designers I've spoken with interact with through Claude Code rather than through Paper.design itself.

But what does this behavior shift mean for how design tools get discovered, adopted, and woven into the fabric of how teams work?

Figma became a juggernaut by building the largest, most passionate community of designers the industry had ever seen. Designers found it, loved it, evangelized it, and enterprise contracts followed. Now, increasingly, the entity choosing the tool is an agent. 

Which forces a question the design tool market hasn't fully reckoned with yet: does community still matter when the agent is the one making the choice? And if it does, how do you build community when the user isn't human?

1. Discovery moves from inspiration to legibility

In the Figma era, designers found tools the same way they found inspiration: stumbling across a showcase, seeing a template they had to try, watching someone they admired post their workflow on Twitter. Discovery was organic, visual, and social. A tool that looked good in a thumb-stopping screenshot could earn its way into someone's workflow that same afternoon.

Agents don't browse for inspiration. When an agent selects a tool for a task, what matters is whether it can understand what the tool does and reliably predict what it will return. Tom Krcha, who runs Pencil, put it plainly when we talked: "People are still building for humans. The bet that's worked for Pencil is being Git-native, with design systems that live in code so coding agents can read them, not in a Figma-only JSON format that agents can't access."

Critically, your MCP server schema, API documentation, and tool registry entry can no longer be the basement window that founders hand over to an engineer. It’s the front door now. 

2. In an agentic world, reliability dwarfs taste

When an agent picks a tool, there's no aspiration in it, no attempt to signal taste. The agent opts for something because it reliably produces good output for a specific kind of task. 

Paper’s rise tells us a lot here. A meaningful share of its actual usage flows through Claude Code, where the agent is reaching for it because it produces the right output for expressive UI: shaders, motion, a particular visual register. In that flow, Paper's brand doesn't mean "beautiful product made by talented designers" (although it certainly has a genuine, passionate human following). It means "the tool that works when you need this specific thing."

The brand that matters most in an agent-mediated world is the job-to-be-done you become synonymous with.

The founders who pick a clear wedge and become the agent's default for that job are building something that compounds. Especially as Anthropic and other large players encroach on the design tooling space, being known for a specific capability is more defensible than being a general-purpose alternative to something bigger.

In addition to reliability, ease of use is paramount. As Anvisha Pai, CEO & Founder of Moda, told me: “The next wave of design tools will do for design what Claude Code did for engineering. The 10x Stripe engineer uses Claude Code to move faster, and so does the vibe coder who doesn’t know what JavaScript is. Design is about to go through the same expansion."

The tools that win will work for the complete beginner while keeping a high ceiling for the expert.

-Anvisha Pai, CEO & Founder of Moda

3. Canvas-first is a dead end for agents

Almost every design tool of the last decade is built canvas-first — it’s where the designer worked, where the work was reviewed, and where the value of the tool was visible. In that architecture, an agent has nothing to grab onto. It can't call your tool, it can't be configured to default to it, and no amount of business development will put you on a list you're technically ineligible for.

The founders who are positioned to win the distribution game are the ones building API-first, where the canvas is one of several surfaces through which humans review and direct the work. Dessn has made this bet explicitly, treating production code as the single source of truth so that a designer's work is accessible to agents from the moment it exists. 

“Design is one of the few non-verifiable reward domains,” Anvisha Pai shared. “Unlike code, there's no objectively good design. For lots of design problems, humans will remain in the loop. The tools that win will be the ones where humans and agents have a level playing field to collaborate. An API-only tool can't do that. A canvas-only tool can't either. The winners merge both modalities.”

Thus, the assertion is not that the canvas will disappear — designers will always want a place to see what their agents are producing and refine the final output. But it should be a window into the product, not the product itself.

4. There are now two communities that matter, and most founders are only building one of them

Community-led growth still works. But the audience has split, and the two communities that matter now require different playbooks.

The first is still the human designer community, though it looks different than the one Figma built. It's smaller, more technical, and shaped more like a design engineer than a pure designer. The leading voices who show up on X, showcase their demos, and gain a following don’t need to pass over their designs to an engineer to translate into code. This evangelist group is still a strong cultural pull.

But the second community is one that most design tool founders haven't started thinking about seriously, and it's where I think the real distribution leverage sits over the next few years. 

Inside platforms like Claude, ChatGPT, and Cursor, the connector and MCP marketplaces are curated lists, and there are real humans deciding what gets featured and what gets surfaced when a user makes a particular kind of request. Inside enterprises, internal platform teams are starting to define which tools their company's agents have access to. And as agents move deeper into the enterprise, that gatekeeping function is only going to become more consequential.

These people are not designers, but being on their radar matters tremendously. 

Figma built a juggernaut by making designers its most passionate constituency. The founders who win the next decade of design tooling will have to do the same thing twice — once for the humans who still shape culture, and once for the agents who are increasingly making the calls.

Special thanks to Tracy Chou, Staff Brand Designer at fal who created the cover art for this piece. I loved this line from Tracy, as it really captures what the next wave of design is all about: “Agents may be increasingly the ones reaching for tools, but a human (me!) is still the one deciding what feels right. I used Fal to generate the images, but the composition, design, and direction were all mine.”