Google Jules
Google's async agentic coding tool that connects to your GitHub repo and handles tasks in the background — no terminal, no local setup, works from any browser.
My take
These days I only use Jules as a monitoring tool: it has a look at my web projects and sends me an email from time to time. That still works. But I've mostly lost interest, and I think so has Google: the changelog has been quiet since March, and Google's coding energy has gone into Antigravity.
At a Glance
Pros
- + Fully browser-based — no installation, no terminal, works on mobile
- + Async workflow means you file a task and come back to a result
- + Connects directly to GitHub repos — no need to clone anything locally
- + A free tier of 15 tasks a day; 100 with Google AI Pro
Cons
- − Async-only — no interactive back-and-forth while it works
- − Less powerful than Claude Code for complex multi-file engineering
- − Tied to Google's models (no choice)
- − Development seems to have stalled: the changelog has been quiet since March 2026
Best for: Scheduled checks on a GitHub repo that run in the background and report back, without a terminal or a local setup
Full review
Jules is Google’s take on agentic coding — but it takes a different approach from most tools in this space. Where Claude Code is interactive and local, and Replit Agent is a full cloud IDE, Jules is deliberately minimal: connect your GitHub repo, describe a task, and come back when it’s done.
The async workflow
The defining characteristic of Jules is that it works asynchronously. You don’t watch it work in real time. You submit a task (fix this bug, check this page for accessibility issues, add this content), Jules runs it in the background, and you get notified when there’s a result to review.
This is either a feature or a limitation depending on how you work. If you like staying in the loop and steering as the agent goes, Claude Code gives you that. If you’d rather fire and forget, Jules fits naturally into that workflow. For recurring tasks like running an automated check every night, the async model is actually ideal.
What it’s good at
Jules works well for bounded, clearly defined tasks:
- Automated checks: Accessibility audits, usability issues, broken links — the kind of thing you’d want running on a schedule
- Content changes: Adding a page, editing copy, updating metadata — tasks with a clear definition of done
- Small fixes: Typos, formatting, minor refactors that don’t require deep architectural reasoning
It’s less suited for complex, multi-step engineering where the requirements shift as you see intermediate results. That’s where interactive tools like Claude Code have the edge.
Mobile-first in practice
One thing that genuinely surprised me: Jules works well from a phone. Because it’s purely browser-based and async, there’s no terminal to manage, no file tree to navigate. You can open your GitHub repo, assign Jules a task, and check back later — all from your phone. For light maintenance work while travelling, that’s genuinely useful.
Pricing reality
Jules has a free tier with 15 tasks a day. The Google AI Pro plan ($19.99/month, the same subscription that gives you Gemini’s paid features) raises that to 100 tasks a day, Ultra to 300. Google currently offers the paid tiers only for personal Gmail accounts.
The agentic coding spectrum
Think of it as a spectrum of friction and power:
| Tool | Setup | Interaction | Best for |
|---|---|---|---|
| Jules | Zero (browser + GitHub) | Async | Scheduled checks |
| Antigravity | Desktop app or command line | Interactive | Visual builds, quick full-stack projects |
| Replit Agent | Zero (cloud IDE) | Interactive | Full apps with hosting included |
| Claude Code | Local (terminal) | Interactive | Complex engineering, full codebase ownership |
Where it stands
Jules’s own changelog has had no new entry since March 2026, while Google has put its coding work into Antigravity, which now has a desktop app, a command line and Claude models alongside Gemini. Jules still runs, and for a scheduled check that reports back it does the job.
See also: Claude Code for when you’re ready for the full interactive agent experience, and Antigravity, where Google’s coding work has moved, for interactive builds.
Added: 2026-03-22 · Last updated: 2026-10-22