Google made several artificial intelligence (AI) announcements on Wednesday. Alongside introducing the new Gemini 2.0 family of AI models, Deep Research feature, and showcasing upgraded capabilities in Project Astra, the Mountain View-based tech giant also unveiled several AI agents. Among them, one agent that can be of particular interest to developers is Jules. The company highlighted that Jules is an AI agent that can create multi-step plans and execute them to solve complex coding-related tasks and uncommon bug fixes.
Google’s Jules AI Coding Agent is Built on Gemini 2.0
In a blog post, the tech giant detailed the new AI agent that can handle coding tasks on behalf of the user. Notably, AI agents are specialised smaller language models (SLMs) that have high efficiency and low latency in carrying out tasks. They can also integrate with specific software to execute various tasks.
Jules is built on Gemini 2.0 AI models and excels in coding-related tasks. Google says it is an experimental code agent which can directly integrate into GitHub workflows. It can create a multistep plan to carry out the actions and execute them. Throughout the entire process, the agent follows the user’s direction and supervision.
Google says that Jules can carry out coding tasks in Python and JavaScript. It can be used to handle bug fixes, code debugging, code reviewing, as well as other time-consuming tasks. The new AI agent can also submit pull requests and land fixes directly into GitHub.
The company claims that Jules can be assigned tasks and can work on them asynchronously. The agent will inform the developer via real-time updates whenever a task is completed, and the developer can also interject to change the priority of tasks. Users can also provide feedback or request adjustments in the agent’s plans to fine-tune its actions.
Jules is currently available to a select group of testers and the tech giant plans to roll out the tool to developers in early 2025. However, to get access, developers will need to sign up for the AI agent in Google Labs.