An autonomous AI coding agent developed by Google. Unlike assistants that offer completions inside your editor, Jules clones an entire GitHub repository into a cloud virtual machine and works through tasks described in natural language without further hand-holding. The typical workflow is to delegate bug fixes, documentation, test additions, or new feature work, and receive the result as a pull request. It launched as a public beta in May 2025 and reached general availability on August 6, 2025.
Key Features
- An asynchronous agent you can hand a whole task to: Given an instruction, Jules clones the repository into a cloud VM, installs dependencies, and then edits files. You can close the browser while it works and pick up the diff as a pull request afterward
- It presents a plan for approval before executing: Rather than editing code immediately, Jules first shows which files it intends to change and how. If the plan misses your intent, you can correct it at that point and avoid wasted diffs
- AGENTS.md carries project-specific conventions: Jules reads an
AGENTS.mdfile placed at the repository root and works with your coding conventions, test commands, and directory layout in mind — so you do not have to explain your team’s style in every prompt - CI Fixer repairs failing checks automatically: Added in February 2026, this feature detects when GitHub Actions checks fail on a Jules pull request and attempts to fix the failure on its own
- External tool integration via MCP: Through the Model Context Protocol, Jules connects to services such as Linear, Supabase, Neon, and Context7, so it can work with issue-tracker and database context in hand
- API access and commit attribution settings: A developer API is available for triggering Jules tasks from external workflows, and you can choose whether pull request commits are attributed to Jules only, co-authored, or to you alone
Pricing
| Plan | Price | Tasks per day | Concurrent tasks |
|---|---|---|---|
| Jules (free) | $0 | 15 | 3 |
| Jules in Pro | Included with a Google AI Pro subscription | 100 | 15 |
| Jules in Ultra | Included with a Google AI Ultra subscription | 300 | 60 |
Paid tiers are not purchased for Jules on its own — they come bundled with a Google AI Pro or Ultra subscription, which also grants priority access to the latest models. Note that paid Jules plans are documented as available only to individual Google accounts (ending in @gmail.com), so Google Workspace accounts cannot subscribe.
Pricing is current as of August 2026. Check the official site for the latest details.
Pros & Cons
✅ Pros
- The asynchronous model lets you fire off a task and move on, clearing chores without stopping your own work
- Even the free tier allows 15 tasks per day and 3 concurrent runs — enough to evaluate it properly on personal projects
- The plan-approval step means you are not handed a runaway diff without warning
- AGENTS.md lets you encode project conventions, so output grows more consistent the more a team uses it
- If you already subscribe to Google AI Pro or Ultra, the higher limits come at no extra cost
⚠️ Cons
- It assumes GitHub integration, so repositories hosted elsewhere are awkward to work with
- Paid plans are limited to individual Google accounts, which raises the barrier for Workspace-centric organizations
- Because everything runs in a cloud VM, tasks that depend on local environment specifics or external service credentials are a poor fit
- As with any autonomous agent, large refactors or vague instructions can produce diffs that drift from your intent
- Someone still has to review the generated changes, so the review burden itself does not disappear
Comparison with Similar Services
| Item | Jules | GitHub Copilot | Devin | Claude Code |
|---|---|---|---|---|
| Provider | GitHub (Microsoft) | Cognition | Anthropic | |
| Where it runs | Cloud VM (async) | In-editor plus cloud agent | Cloud VM (async) | Local terminal |
| Main entry point | Web browser | Editor / GitHub | Web / Slack | CLI |
| Integrates with | GitHub | GitHub | GitHub and others | Any local repository |
| Free tier | Yes (15 tasks/day) | Yes (limited) | No | No (bundled with subscription) |
The entry points differ: Copilot for completions while you type, Claude Code for working through a local repository conversationally. Jules suits work you can carve out and hand off entirely.
Who Is It For
- Developers sitting on a backlog of test additions, documentation cleanup, and minor bug fixes
- Teams that manage repositories on GitHub and already have a pull-request review culture
- Anyone already subscribed to Google AI Pro or Ultra who wants to try a coding agent at no extra cost
- People who want an agent working in an isolated environment rather than on their local machine
- Anyone who wants to measure an autonomous agent’s real capability on a free tier before committing
Summary
Jules commits fully to one way of working: hand over a task against a GitHub repository, get a pull request back. Plan approval and AGENTS.md give you control over the direction of its output, and the free tier makes everyday experimentation easy. On the other hand, its GitHub dependency and the individual-Google-account requirement rule it out in some environments. Start on the free plan with a few small tasks and see whether it fits your workflow.