Agenvas

By the time a short AI video is done, it lives in one chat window and five browser tabs, with the voice-over somewhere in a downloads folder called "final2". Agenvas puts the whole project on one large board that you install on your own computer or server. Every piece of text, image, video or sound becomes a card, and lines between cards show what was made from what.

The AI work runs on services you connect with your own API keys, and each provider bills you for what it generates. Text works with any provider that follows OpenAI's interface, while media goes through ComfyUI, RunningHub and a few named models. A built-in agent reads the cards linked to it and rewrites text, but any media it generates appears first as a proposal you accept or turn down.

Installing takes one command for Docker, a program that runs packaged apps. The code is on GitHub under the MIT license, with release numbers still at 0.0.x. The site says API keys are encrypted at rest and kept out of logs. Each install has a single admin account, which also decides which models the other users see.