Genvid’s production system is now something your own agent can drive. An MCP server, a CLI, and a skills pack, in beta for every subscriber.
Jerry Heinz, CIO, Genvid
Every day I build software with a coding agent. It’s increasing fast and robust, but that’s not the thing that keeps me coming back daily. There is a sense of pure, unadulterated joy that comes from being able to experiment, prototype, and explore paths just to see where it goes or if I can do it. I can ask it to build a small tool that exists only because I work a certain way. Over time my harness has come to understand how I think and what makes my approach … me.
We built Genvid to solve the inherent problems of scaling AI-based TV and film production. Our system applies governance and unifies your workflow so entire teams can collaborate on a project, while knowing their content is safe, secure, and tracked.
However, I kept thinking that a producer, a director, an artist deserves that same feeling of joy. A creator simply wants make art. The work they resent is the everything around it. With most tools (AI-based or not), tracking what is approved, proving where an asset came from, carrying one note across a hundred keyframes, and finding that single render in a mountain of assets is the drudergy that bogs us all down. A capable agent should remove that cognitive load, so the artist is free to spend their time doing what they love.
Today we open the beta that brings that experience to creators.
Bring your own agent to production
Genvid is the production system for generative AI video. It runs on the production data model studios already use (project, scene, shot, asset, task), aligned with the MovieLabs Ontology for Media Creation, and it captures provenance the moment each asset is made.
What’s changed is you reach it: Genvid now fully exposes its production workflow as a governed boundary that your agent reaches through an MCP server, a command-line interface, and a reference skills pack. You drive that boundary from the agent you already use (Claude, Codex, Cursor, OpenClaw, Hermes, etc).
Generate anywhere. Govern in one place.
Your agent learns your studio. The house rules live with it as skills you author. It can explore many candidates for a shot, carry a change once you decide to make it, and answer where a project stands, all against the same governed graph that holds your approvals and your budget. Genvid handles provenance and governance underneath, so the creative work stays unencumbered.
The same context that makes governance possible also makes an agent better. An agent working against a clean, attested, structured record produces more consistent and more efficient results than one working blind. Good provenance keeps the work traceable and attributable, and it makes an agent reliable enough to trust with professional production work.
Getting started
The beta is open to every Genvid subscriber. To get started, point your agent at our public GitHub repo and tell it to install. Then start talking with yout agent about your production and see how it feels to explore again.
We have believed one thing for a long time. The platforms that survive the agentic shift are the ones that give agents a data model, an open surface, and a provenance chain to reason against. We built Genvid that way on purpose.