In this video I tested Google’s new autonomous coding agent, Jules, and it honestly feels like having a whole team of developers working on my GitHub repo while I’m asleep. Jules connects directly to your GitHub org or a specific repo, spins up an isolated cloud VM on Google’s backend, clones your repo, and then uses Gemini 2.5 Pro to execute tasks end-to-end: write code, run unit tests, debug, and finally open a pull request for you. The key difference is it’s fully asynchronous and remote—nothing is running on your local machine unless you want it to. I walked through a real demo where I started from an empty repo and launched three tasks in parallel (UI, backend, and features) to show how Jules scales work across multiple sessions/branches. Jules even launched the app in the VM, took screenshots to verify behavior, and iterated on fixes before asking me to create a PR. I also covered how to run Jules from the CLI (npm install jules, then jules remote …), and why the new Jules API is a big deal: you can wire it into MCP, Slack/GitHub issue automation, Jira tickets, or CI/CD so agentic work happens automatically. Finally, I broke down pricing/free limits (15 tasks/day free, 3 concurrent; up to 60 concurrent on Ultra) plus features like env vars and memory.