Is Jules Free? Google AI Coding Agent Plans (2026)
Google's Jules agent runs coding tasks unattended in the cloud, drafting a plan before it edits code, and opens a pull request for a developer to review.
Jules suits developers who want bug fixes and small features handled in the background instead of blocking their day, not a live pair programmer for real-time back and forth. During its public beta, users shipped over 140,000 code changes through it, and it now replaces manual backlog triage for teams drowning in minor GitHub issues.
Jules is Google's asynchronous AI coding agent, generally available since August 6, 2025, that clones a GitHub repository into an isolated Google Cloud virtual machine, plans a fix, and opens a pull request for human review, running on Gemini 3.1 Pro.
Maker: Google · Autonomy: semi autonomous · Maturity: GA
Underlying models: Gemini 3.1 Pro, Gemini 2.5 Pro
About Jules
Jules is an autonomous coding agent built by Google Labs, publicly launched in beta at Google I/O in May 2025, and made generally available to everyone on August 6, 2025. Unlike a chat-based coding assistant that answers questions inside an editor, Jules receives a full task, works on it unattended in an isolated cloud environment, and comes back with a pull request ready for review, closing the loop between reporting a bug and shipping the fix. Under the hood, Jules clones the target GitHub repository into a dedicated Google Cloud virtual machine, reads the codebase, drafts an execution plan, and only proceeds once that plan is visible to the developer. It originally ran on Gemini 2.5 Pro and now defaults to Gemini 3.1 Pro for Google AI Pro subscribers, giving it long-context reasoning across large repositories. Because every action happens inside its own VM rather than the developer's live environment, Jules can run tests, install dependencies, and rewrite files without any risk to a local machine. Jules works best on specific, well-defined tasks such as fixing a named bug, upgrading a dependency, or writing missing unit tests, rather than vague requests like 'the login is broken.' It suits backend engineers clearing a bug backlog, open-source maintainers triaging community-filed issues, and small teams that want routine dependency and lint fixes handled without a human opening a laptop. Google Workspace and enterprise Google accounts cannot subscribe to a paid Jules tier as of mid-2026; only individual @gmail.com accounts can upgrade. Access runs through the jules.google web app, the Jules Tools command-line interface, and an alpha-stage Jules API for wiring tasks into Slack, Linear, or CI/CD pipelines like GitHub Actions. Google keeps shipping Jules on a roughly weekly cadence: CI Fixer automatically resubmits a pull request after a failing GitHub Actions check without a human re-prompting it, Scheduled Tasks let developers set up recurring jobs like a nightly lint pass, and Suggested Tasks scans a codebase for #TODO comments and proposes a plan on its own initiative.
Pricing
Free: 15 tasks/day, 3 concurrent tasks. Google AI Pro: $19.99/mo for roughly 100 tasks/day. Google AI Ultra: $124.99/mo for roughly 300 tasks/day. A $100/mo and $200/mo developer Ultra tier (5x/20x Pro usage) rolled out in May 2026. Only individual @gmail.com accounts can subscribe; Workspace and enterprise accounts are unsupported as of mid-2026.
Key Features
- CI Fixer: Detects a failing GitHub Actions check on a pull request it opened and resubmits a fix automatically, without a human re-prompting it.
- Scheduled Tasks: Runs recurring jobs on a set cadence, such as a nightly lint pass or a weekly dependency check, without a developer kicking it off each time.
- Suggested Tasks: Scans a connected repository for #TODO comments, drafts an execution plan around them, and surfaces it for approval.
- Isolated Cloud VM Execution: Clones the full repository into its own Google Cloud virtual machine per task, so it can install dependencies and run tests without touching a developer's local machine.
- Jules API and CLI: An alpha-stage REST API plus the Jules Tools command-line interface let teams trigger and monitor tasks from Slack, Linear, or a CI/CD pipeline.
Strengths
- Runs entirely asynchronously, so a developer can hand off a bug fix and keep working instead of babysitting a live session.
- CI Fixer automatically detects a failing GitHub Actions check on its own pull request and resubmits a corrected version without a human re-prompting it.
- Free tier requires no payment method and works out of the box across Python, TypeScript, JavaScript, Go, Rust, and Java repositories.
Weaknesses
- The free tier's daily task quota and concurrent task cap are thin for a team clearing a large bug backlog.
- Google Workspace and enterprise Google accounts cannot buy a paid Jules tier as of mid-2026, blocking a company-wide rollout.
- Struggles with vaguely worded issues such as 'the login is broken' and can return verbose code that needs manual trimming.
Frequently Asked Questions
How much does Jules cost in 2026?
Jules is free for individual Google accounts, with 15 tasks per day and 3 concurrent tasks. The Google AI Pro plan costs $19.99 a month and raises the limit to roughly 100 tasks per day, while Google AI Ultra costs $124.99 a month for about 300 tasks per day. A separate $100 and $200 monthly developer Ultra tier (5x and 20x Pro usage) rolled out in May 2026, but Google Workspace and enterprise accounts still cannot subscribe to any paid tier as of mid-2026.
Is Jules fully autonomous?
Jules plans and executes coding tasks with minimal supervision, but it is semi-autonomous rather than fully hands-off: it proposes a plan before writing code and always stops at a pull request for a human to review and merge. It cannot merge its own pull requests or push directly to a production branch. CI Fixer lets it resubmit a pull request automatically after failed CI checks, narrowing the loop but keeping a human as the final gate.
What are the best alternatives to Jules?
Devin from Cognition targets longer, more autonomous engineering sprints and skips Jules's plan-approval step, better suited to teams comfortable with less oversight. GitHub Copilot Workspace stays closer to the GitHub UI for teams that want agent tasks without leaving the pull request review screen. Claude Code is a better fit if you want an interactive terminal agent for live pair programming rather than Jules's fire-and-forget async model.
How does Jules compare to Claude Code in 2026?
Jules runs entirely asynchronously in a Google Cloud VM and only surfaces once it has a finished pull request, while Claude Code works interactively in a terminal in real time. Jules is free up to 15 tasks a day and scoped to GitHub-based Python, TypeScript, JavaScript, Go, Rust, and Java projects, while Claude Code supports live back-and-forth across a broader range of workflows. Choose Jules for background bug-fix batches, Claude Code when you want to steer the agent step by step.
How do you get started with Jules?
Sign in at jules.google with a personal @gmail.com Google account, connect a GitHub repository, and assign Jules a specific, well-scoped task such as fixing a named bug or adding a small feature. Jules clones the repo into an isolated VM, proposes a plan for approval, then executes it and opens a pull request, typically within minutes for small tasks. The free tier requires no payment method, so anyone can try it before upgrading.
Top Alternatives
- Claude Code: Pick Jules if you want async background pull requests; pick Claude Code if you want an interactive terminal pair programmer.