Jules Freemium

What it is

Jules is Google's asynchronous coding agent. You connect a GitHub account, pick a repository and branch, and describe a task, or label a GitHub issue with "jules". Jules clones the repository into a Google Cloud virtual machine, writes a plan using Gemini models, shows you the plan and diff for approval, makes the changes and opens a pull request. Because it runs in the cloud, you can hand it work and carry on with something else.

Key features

• Plans before coding, with the plan and diffs available for review
• Runs in an isolated cloud VM, so nothing runs on your machine
• Trigger tasks from the web app or by labelling GitHub issues
• Optional environment setup scripts so the VM can install dependencies and run tests
• Jules Tools, a command-line interface installed with npm install -g @google/jules, for starting and pulling remote sessions from the terminal
• A REST API for integrating Jules into your own workflows

Skills, subagents and MCP

Jules automatically reads an AGENTS.md file in the repository root to understand the codebase and conventions, which improves its plans. Its documentation does not cover Agent Skills, subagents or MCP servers, so project context should go into AGENTS.md and your setup scripts.

Pricing

Jules has a free tier and higher limits on Google AI subscriptions. At the time of writing (September 2026), the Jules site lists 15 daily tasks and 3 concurrent tasks on the base plan, 100 daily and 15 concurrent on Jules Pro, and 300 daily and 60 concurrent on Jules Ultra, with the paid tiers getting newer Gemini models. Pro and Ultra come with Google AI Pro and Ultra subscriptions; check the Jules site and Google's plan pages for current prices and student offers.

Who it is for

Developers on GitHub who want to offload routine work such as bug fixes, version bumps, tests and small features, and who prefer reviewing a pull request to watching an agent in a terminal.

Verdict

Jules is one of the easiest cloud agents to try thanks to its free daily tasks, and the plan-then-PR workflow is well suited to routine changes. It is GitHub-only, less configurable than CLI agents, and without documented MCP support it cannot reach external tools the way some competitors can.