Kino AI launches collaborative video editing with 15+ agents on a single timeline

Video editing works poorly with AI agents not because models cannot write scripts or assemble animations. The problem is more mundane: source files sit on drives, the edit lives in a separate application, feedback arrives as screenshots in Slack, and the relevant segment has to be described in words. Kino AI proposes bringing all of this into a single addressable project where people and agents work on a shared timeline.

A simple example: a product manager asks to replace a shot at the 42-second mark of a promotional video. In a conventional workflow, they send a screenshot, write a comment, and hope the editor or agent finds the correct layer. In Kino, a specific frame, clip, comment, and even line of code has its own link. That link can be passed to Codex or Claude Code, giving the agent an exact object to modify instead of a verbal puzzle.

The edit becomes a state accessible to agents

Kino is an editor in which the timeline, graphics, element movements, and 3D scenes are represented as Remotion-based code. At the same time, users see a familiar visual interface and can move elements manually. Changes are synchronized in real time: participants can see one another's cursors and the section of the video a colleague is currently viewing.

To work with recorded footage, the service transcribes speech, builds vector representations, describes the content, and recognizes faces. The local file remains available for fast previewing, while a lightweight copy is uploaded to the cloud for colleagues and models to work with. The developers say the editor accepts professional ProRes, BRAW, and R3D formats, and that a 250 GB ProRes file can be dragged directly into a project. Playback is handled by custom decoders, WebCodecs, and Mediabunny.

The practical value is not another “make a video” button. Kino is trying to give agents what media production usually lacks: a single source of truth for the current state and precise links to objects. It resembles collaborative software development, where a team works in one repository and discusses specific lines instead of exchanging archives.

What already looks useful

According to the founders, some users run 15 or more agents on a single project: one handles rotoscoping, another performs color correction, and the rest assemble individual segments. Teams also use links to review results: a person leaves feedback on the relevant frame, and an agent applies the correction to the shared timeline.

MCP access is particularly interesting for developers. A custom agent can be connected to a project and assigned editing operations without a separate, custom-built interface. The basic editor, including import, export, collaboration, and MCP, is available for free. The $25-per-editor-per-month Pro plan adds unlimited MCP tool calls, more storage, shared links, and YouTube search.

Where caution is still warranted

The article was published by Kino's creators, so accounts of 15+ agents working together and faster production cannot be treated as an independent comparison. There is no data on export times, stability on large projects, the quality of automatic color correction, or compute costs. It is also unclear how conveniently conflicts can be resolved when several agents modify the same scene simultaneously.

Privacy is another concern: a lightweight copy of the source material is uploaded for cloud analysis, and the service performs face and content recognition. Teams working with confidential footage should review the data retention, deletion, and usage policies in advance. Kino is currently available only as an Electron application; there is no browser version.

The main takeaway for product teams is that Kino's value should be tested not by the visual appeal of its automatically generated graphics, but by the quality of the shared project state. Take a real video, connect two agents and one editor, then measure the time from comment to accepted revision, the number of incorrectly modified segments, and the number of manual rollbacks. If the addressable timeline genuinely reduces those losses, this is more than an AI feature: it is a new working layer for video production.