Google has released its latest agentic large language model (LLM) tool, Jules, designed for coding tasks and collaboration. We tested Jules on a simple Rails CRUD project with Bootstrap integration, comparing it to Claude Code.
Jules clones your repository into a private virtual machine and starts working, producing a formal privacy notice and a 90-second video tutorial. The AI generates plans based on code snippets provided by the user, making suggestions for improvement. This approach is appealing for non-developers, but it may require users to understand Git semantics.
To get started with Jules, we pushed our example project to GitHub and created a repository. We then connected Jules to the repo, providing code snippets for tasks like applying Bootstrap definitions and updating buttons.
The results were promising, as Jules cleaned up the sidebar in our Rails app, making it more visually appealing. However, some minor improvements could have been made with more specific prompts. The process is slightly longer than using Claude Code, but this can be attributed to the use of pull requests.
While Jules may not have matched Claude’s level of intent understanding, its internal framework recognizes the growing causal coder market. Google’s efforts in providing an alternative coding solution on their own hardware are notable. Despite some inconsistencies, Jules appears to mark a significant step forward in agentic coding tools.
Source: https://thenewstack.io/agentic-coding-how-googles-jules-compares-to-claude-code