Kino launches from closed beta: a Figma-style collaborative video editor built for AI agents and human teams

Kino is a browser-based collaborative video editor built to let human teams and AI agents work simultaneously on a shared timeline. Luke Igel, the company's founder, launched it publicly this morning after an extended closed beta.

The Figma-for-video pitch

The core idea is that video production is inherently iterative and team-based, which makes the dominant tools, desktop-native apps like Premiere and Blackmagic Design, poorly suited for the way AI-augmented teams now work. Kino puts everyone, people and agents alike, inside one shared timeline. Igel and a colleague recently spent several days building a launch video that way, with multiple AI agents running in parallel against the same project, Figma-style.

The product is positioned as a managed service rather than a bare orchestration layer. Igel wants users to bring their own AI subscriptions, OpenAI, Claude, whatever they're already paying for, and run those through Kino rather than having the platform charge for model access. His argument is that if a team burns through $10,000 in Claude tokens doing real video work, those tokens should be spent inside Kino rather than scattered across disconnected tools.

“We are giving that a home, in all of the things that you need to do real work where all of your teammates' footage and all of your teammates' edit decisions are in one place. We spent the past few days living in one giant editor timeline, Figma style, while swarms of my agents and swarms of his agents all just chipped away at the same mega timeline.”

The technical bet on WebCodecs

Kino is built on WebCodecs, a browser API that enables GPU-accelerated video processing roughly comparable to FFmpeg running natively. Igel is direct that this business could not have been built in 2024; the infrastructure only became viable recently. The practical payoff is scrubbing performance that feels competitive with native apps, which Igel treats as a hard prerequisite for user retention.

The team has also built custom decoders for formats that don't resolve cleanly in a browser context, including ProRes, Blackmagic RAW, and footage from older consumer camcorders. That codec coverage matters because without it, agents hitting those files would fail silently, and the long tail of format problems is part of what Kino is managing on the user's behalf.

LLMs over diffusion

On generative video, Igel is skeptical of the pure text-to-film model and explicit about preferring LLM-driven production. Given a choice between a diffusion model that generates a Pixar-style film from a prompt and a coding agent that directs Blender and RenderMan to produce the same result, he would take the agent. Controllability is the reason. The launch video Kino used to demo the product mixes real footage, motion graphics, rotoscoped screen replacements, and a generated shot from a new C-Dance model checkpoint the team obtained access to just before shipping. None of those layers are distinguishable at a glance, which is the point.

His current workflow skips prompting almost entirely. He feeds a Granola meeting transcript of himself and a collaborator discussing what the video should be, and Kino uses that context to lay out a rough cut alongside a storyboard. The agent isn't replacing creative decisions; it's executing on decisions that were made conversationally.

Where things stand

Growth figures weren't disclosed. Kino has been in closed beta until today's launch, so the public numbers don't yet exist.