YouTube Automation Agent is an open-source, self-hosted AI agent that runs a YouTube channel end to end. It researches topics, writes scripts, generates narration, visuals and finished videos, optimizes them for search, and publishes them after your approval. The Node.js project has more than 4,400 stars on GitHub and is actively developed.
The agent is organized as a full content pipeline rather than a single script. A research stage spots topics and angles worth covering; a writing stage produces scripts; production stages synthesize voice narration, gather or generate visuals and thumbnails, and assemble the final video, including Shorts. Before anything goes live, the creator reviews and approves the upload, keeping a human in the loop. After publishing, it reads channel analytics and feeds the lessons back into future topic and title choices. It works with several model providers, including Google Gemini, OpenAI, OpenRouter and DeepSeek, so you can balance cost and quality per stage.
Configuration is handled through environment files, so a channel can switch model providers, voices and visual styles without code changes. The codebase is modular, and individual stages such as scripting or thumbnail generation can be lifted out and reused in other content pipelines.
The agent is free and open source under the MIT license and self-hosted. Your only expenses are the model and media API usage you configure; there is no subscription or revenue share.
YouTube Automation Agent is for creators and small media teams who want to scale a channel’s output with AI while keeping editorial approval, and for developers interested in a complete, real-world agent pipeline they can adapt.