A release workflow
Tag a version, and the same job that builds the binary renders the video that announces it. The CLI exits 0 or 1, so a failed render fails the build rather than shipping a stale clip.
video as code
Compositions your agent writes and re-renders, on a timeline you can watch. Deterministic local rendering, and project files that sit in git next to the product they demonstrate.
macOS · Windows · Linux · no credits, no render limits
why it works
Coding agents are good at one loop: read, edit, run a command, check the result. A CoAnimator project is built for that loop: everything the agent needs to touch is an ordinary project file, and the editor hot-reloads each edit onto the stage before the agent's next turn.
# render a project headlessly, same pipeline as the editor
coa render my-demo --preset 1080p60 --runners 8
# a packaged project file works too
coa render launch-video.coa --out launch.mp4setup
The desktop app ships the coa command line tool, and one menu click symlinks it onto your PATH. A second installs the coa skill, so your agent already knows the flags, the presets and the project layout.
Template projects seed their own skills: install one and the agent learns its looks with it.
determinism
Every frame is a pure function of the timeline clock, so the exporter can render any moment on demand and the result is reproducible run after run.
coa render runs the same chunk-parallel export pipeline as the desktop app, so headless output matches what you approved on the stage, at the same settings.
Renders run on your hardware, laptop or CI runner, with unlimited exports and no per-export fees. Nothing is metered and nothing uploads.
video as code
The whole video lives as project files: they diff, they review, they sit in git next to the product they demonstrate. That is what makes videos maintainable, since a UI change is a pull request rather than a re-shoot, and it is what makes scheduled video possible: a cron job, an agent reading the week's merged PRs, and coa render at the end.
in the pipeline
Tag a version, and the same job that builds the binary renders the video that announces it. The CLI exits 0 or 1, so a failed render fails the build rather than shipping a stale clip.
Nothing has to be open. A scheduled run reads the week, updates the project and renders it, which is how a weekly changelog video gets made without anyone remembering to make it.
The project diffs like code, so a change to the demo is reviewable next to the change it demonstrates. Approve both together and the video is never a version behind the UI.
The same command, the same pipeline, the same output. There is no cloud tier that renders differently from the machine you tested on.
Yes. The coa CLI renders CoAnimator projects to MP4 without opening the app, so exports can be scripted and run in CI. It exits 0/1, takes a project id, a folder, or a .coa file, and runs the same local, chunk-parallel pipeline as the desktop app with presets from 720p30 to 4K60.
Through skills, not a proprietary API. Templates ship with skills that teach Claude Code, Codex, Antigravity, or OpenClaw each format’s rules, and the editor reflects the agent’s changes live. Any AI agent you already use can drive the studio.
It’s adjacent but different: Remotion is a React framework where developers program videos; CoAnimator is a desktop studio where agents author videos on a visual timeline, with local rendering built in. If you want a video product built in React, Remotion is great; if you want finished videos produced by agents with a human directing, that’s CoAnimator.
CoAnimator renders MP4 video today, in-app or via the coa CLI. Lottie export is on the public roadmap but not shipped yet.
They run wherever you can install the app or the CLI, which includes a CI runner or a server you control. What there is not is a hosted render service you submit jobs to: rendering is local by design, which is what keeps it unmetered and keeps your footage off hardware you do not own.
The desktop app ships native Windows builds and renders locally there; the packaged coa shim installer is macOS-first for now, with the same render pipeline available in-app.
Install CoAnimator, point it at the agent you already pay for, and describe the video. Every project stays editable and renders on your own machine.