How Endless rebuilt its AI creative workspace with tldraw
Endless is an AI platform for creative teams, built with tldraw’s SDK. Its canvas is where the team works, and where Endless's agents are starting to work too.
Challenge
Endless is a multi-AI platform that brings a curated set of leading AI models together under one subscription, so a whole team can research, write, generate images and video, and review work in the same place. It has been canvas-first since it started in 2023.
The bet, from Francesco’s first pitch, was direct and collaborative manipulation: creative work with AI needs a surface people can touch, where they drag, group and rearrange what they're making, with room for images, drafts and ideas — all of which would be an unworkable nightmare over a stream of messages.
They built that canvas themselves first, on React Flow, Konva.js and the browser's own canvas APIs. Fairly soon, however, the strain started to show. Paste more than ten shapes on the board and things got laggy. The app broke in Safari. There was no undo or redo. By 2023, about 80% of the team's time went into keeping the canvas alive instead of building the product around it. When customers started asking for collaboration, the team scoped multiplayer on that stack (sync, presence, everything needed for a team to work on the same board) at least six months of engineering.
In 2024, looking for canvas infrastructure that could meet their needs and ambitions, they found tldraw's SDK through Steve’s experiments on X. Endless shipped an early version on it and ran that commercially for over a year before rebuilding the whole product from scratch.
Solution
Within a week of adopting the SDK, Endless had every canvas feature it needed running, sync and multiplayer included, against the many months & quarters it had scoped on its own. When the team rebuilt the whole product from scratch they were able to keep the canvas at the center.
"What we are doing would be almost impossible without tldraw," — Francisco Parata, Endless's co-founder and CEO.
Endless has built well past the canvas since then — into chat, content generation, social research — and the canvas has become the enabling factor for a lot of Endless’ latest features, with more to come.
Part of that freed-up time goes into putting agents on the canvas. Endless's founding thesis is the power of direct manipulation. Chat pushes people into linear thinking, one message after another. A canvas lets a team branch out, work in parallel and converge again, exactly how creative work happens.
"Technologies for the future need interfaces from the future," Francisco says. "We think canvas is one of those interfaces."
For agents, the canvas has to be more than a place to draw on. tldraw's SDK exposes that canvas as an API, the same interface whether a person or a script moves a shape, and that is what lets Endless build an agent that acts on a board instead of describing it. In an upcoming release, you can ask Endless's sidebar assistant to organize a board by color and it does the work for you: red images into one group, blue into another, yellow into a third, with its cursor moving across the board as it goes. Endless built that agent's behavior. tldraw's SDK is what made the canvas something an agent could access at all.
Results
Endless users are able to collaborate in a way they simply couldn’t before. Not just that, but Endless also runs in-person AI innovation days on its own product that have drawn countless participants across different editions, with up to 120 people working on the same board at once. The canvas held up.
Everything is now in one place. A team using Endless stops moving between a chat tab, a drive folder and a design file. References, generations, and drafts live on a single board, with comments on the canvas in development, and everyone looks at the same version at the same time. Nobody has to re-explain where anything is before it can be reviewed.
Endless's next move is to put more of its agents to work the same way, acting directly on the canvas instead of describing changes to it. For teams starting to build agents on the same pattern, tldraw's agent starter kit is built for exactly that pattern and is a natural next stop for teams already building agents that move shapes for a living.
Want to know more? If you're ready to build your own canvas, get started with the tldraw SDK.
About the author

Dalila Margiotta
Content Marketing Manager at tldraw
Dalila Margiotta is a Marketing Manager at tldraw, where she spends her days finding interesting stories and getting more people to care about them. She's endlessly curious about technology, creativity, and people building things that didn't exist before. Outside of work, she's usually bringing ambitious people together, dreaming up her next project, or convincing someone to turn a slightly ridiculous idea into a real one.