Jules
Jules is Google's coding agent, which works independently in the cloud and in the background on software code, then presents the result to you for approval.
In plain terms
Jules is like a subcontractor to whom you hand over the house keys and a task list. You go to work, while it paints the room and shows you what it did in the evening. You only accept the work after reviewing it and being satisfied.
In practice, a programmer gives it a task in text, for example 'fix this bug' or 'write tests for this part'. Jules then works not on the developer's machine but in Google's cloud, on an isolated virtual machine, on a copy of the code.
This differs from assistants built into code editors that suggest as you type. Jules is asynchronous: you do not need to sit and watch it; you can assign multiple tasks at once and check back when they are done.
How it works
- You connect Jules to your code repository (typically on GitHub) and describe in plain language what you need.
- Jules copies the code to an isolated virtual machine running in the cloud and uses the Gemini model to read it and understand the project.
- It creates a plan of steps that you can review, modify, or approve before it gets to work.
- It executes the plan: it modifies files, runs commands and tests, and if something fails, it tries again.
- At the end, it shows the changes and opens a pull request, which is decided by a human.
An everyday example
On a Friday afternoon, the developer of a small online store asks Jules to update outdated software libraries and write tests for the shopping cart. On Monday morning a finished proposal is waiting: the developer reviews it, sends one item back, and merges the rest with a single click.
Who is affected
A common misconception
Limits and risks
- It can be mistaken: the code may run and tests may pass even if the solution is flawed or contains a security vulnerability, so review cannot be skipped.
- Your code ends up in an external company's cloud, so for a confidential or regulated project, first clarify whether this is permitted.
- Vague requests yield vague results; the more precisely you describe the task, the more useful the suggestion will be.
- The number of daily tasks is limited depending on the package, and a larger quota requires a subscription.
Similar, but not the same
- Gemini CLI – The Gemini CLI typically works on your own machine, in the command line, in conversation with you. Jules, however, runs in the cloud, in the background, without requiring your attention; you verify the result at the end.
- Code completion – Code completion suggests a few lines while you type, while Jules carries out a full task from planning to the merge request.
- Gemini – Gemini is the language model itself, while Jules is a product built on it that provides tools and a work environment for the model.
Related concepts
coding agent, AI-agent, Aszinkron agent
Jules Tools
Gemini CLI, Claude Code, OpenAI Codex, GitHub Copilot, Devin
A brief history
Google introduced Jules as an experimental project in Google Labs in December 2024. It entered public beta at the Google I/O conference in May 2025 and became generally available in August 2025.
Where things stand
Jules has been a generally available product since August 2025, with free and paid packages, a command-line tool, and a programming interface. It is one of the well-known players in the crowded field of asynchronous coding agents, but for most developers it is still not a daily work tool.
In the news
- Google shuts down Jules coding agent; data export open until November 11Oct 30, 2026 · Code
Also known as
Jules · Google Jules · Jules agent · Jules AI · Jules coding agent · Jules coding agent · Google's coding agent · Jules Tools
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.