Forget prompt-to-video toys. Mirage is going after professional post-production workflows, and it’s doing it by making AI agents the editor. The company announced Tesseract, a video creative suite built specifically for AI agents to operate, not humans clicking through timelines.
Tesseract gives agents direct access to a video engine built around layers, compositions, keyframes, masks, adjustment layers, timing, and audio. That’s a meaningful technical distinction. Most current agent-based video tools either generate short clips from text or stitch together pre-rendered assets. Tesseract lets an agent actually construct and edit a project, the way a human editor would in Premiere or After Effects, but through code and instructions rather than a GUI. CEO Gaurav Misra put it plainly: “Premiere and After Effects, rebuilt for agents.”
The use cases span product launch films, social ads, animated typography, and footage with graphic overlays. Users supply assets and creative direction, the agent builds and iterates, and the project stays editable throughout. Agents can adjust a specific keyframe, retime a title, or rebalance audio on request. Finished videos can be exported, and projects can be reopened for revisions later.
The whole engine runs locally for free. No Captions account, no API key, no credits. Users only pay for whatever agent or model they’re already using. That pricing model is a smart distribution play: Tesseract slots into existing workflows rather than asking people to adopt a new platform.
This fits a broader shift where AI infrastructure is moving closer to production-ready creative work. Runway, Pika, and Adobe’s Firefly Video are all pushing into similar territory, but none of them have opened a local, agent-native editing engine like this. Tesseract’s installation is on GitHub, and the local engine supports any compatible agent environment with file and tool access.
