2026 Review

Jules

Google's asynchronous coding agent — hand it a task on a GitHub repo and it comes back with a PR.

CodingFreemiumAPIPopular78% positiveAgent

Facts checked Aug 9, 2026

About Jules

Jules is an async coding agent from Google: you point it at a GitHub repository and describe a task, it clones the repo into a cloud VM, writes and tests a plan, and opens a pull request for you to review. There is nothing to install and no editor to sit in — you delegate and come back. The free tier runs 15 tasks a day with 3 in parallel on Gemini 2.5 Pro; Google AI Pro at $19.99/month raises that to 100 tasks a day on the newest models, and Ultra to 300.

At a glance

Founded: 1998
Headquarters: Mountain View, US
Platforms: Web

Integrates with

GitHub

Main competitors

Codex Devin GitHub Copilot Cline

Best for

Bug fixesDependency upgradesTest coverageCode cleanups

Pros

A genuinely free tier of 15 real agent tasks a day, which is more than most agents give you at any price
The async model fits the work: small, well-specified chores get done while you are doing something else
Output is a pull request against a branch, so nothing lands without review
Parallel tasks mean you can fire off five small fixes at once instead of queueing them

Cons

Async is the wrong shape for exploratory work — you cannot steer it mid-task the way you can Cursor or Claude Code
Paid access is bundled into Google AI subscriptions rather than sold on its own, so you buy a lot of unrelated product to raise your limits
Upgrade paths are documented for individual Gmail accounts; Workspace and business access remains restricted
Task quality falls off sharply on vague instructions — it needs a well-specified job, not a vague direction

Our verdict

Best for: Developers with a backlog of small, well-defined tasks — dependency bumps, test coverage, tidy-ups — who want them done in the background.

Skip if: You want a pair programmer to think alongside you; an interactive agent like Claude Code or Cursor is a better fit for exploratory work.

Our review

Jules bets on a single idea: most of what a coding agent is good at does not need you watching. Dependency bumps, a failing test, a rename across forty files, the tidy-up you have been postponing — these are well-specified jobs with a clear finish line, and Jules takes them, works in a cloud VM, and comes back with a pull request. Reframing agentic coding as delegation rather than pairing is the most interesting thing anyone has done with the category this year.

It works best when you treat it like a junior with good instructions. Give it a precise task and the result is usually a mergeable PR; give it "improve the error handling" and you get something plausible and unfocused. The plan-first behaviour helps a lot — it tells you how it intends to proceed, and correcting a plan costs a sentence where correcting a diff costs an afternoon. Parallelism is the underrated part: three concurrent tasks on the free tier means the backlog clears in one pass rather than five.

The free allowance deserves particular note. Fifteen real agent tasks a day, no card, is more generous than anything comparable — Devin costs real money before you learn whether it suits you. The catch is upward: paid limits are not sold on their own, they come attached to a $19.99 Google AI Pro subscription that also includes Gemini, Flow credits, storage and YouTube Premium Lite. If you want that bundle anyway, Jules is nearly free; if you want only Jules, you are paying for a lot of things you did not ask for. Workspace accounts also remain awkwardly served.

Verdict: the best async coding agent available and an easy recommendation next to an interactive tool rather than instead of one. Keep Claude Code or Cursor for thinking work, and send Jules the chores.

What reviewers elsewhere say

78% positive — Developers like the free allowance and the fire-and-forget workflow for small chores; the recurring criticisms are weak results on vaguely specified tasks and having to buy a whole Google AI subscription to raise limits.

Sentiment estimated from reviews on Reddit, Product Hunt, as of Aug 2026. Third-party scores are point-in-time snapshots. We don't host user reviews.

Key features

Asynchronous by design — tasks run in a cloud VM while you do something else
Direct GitHub integration: clones the repo, works on a branch, opens a pull request
Writes and shows a plan before it edits, so you can correct the approach rather than the diff
Runs several tasks in parallel — 3 on Free, 15 on Pro, 60 on Ultra
Public API for triggering and managing tasks programmatically
Runs on Gemini 2.5 Pro free, with priority access to the newest Gemini models on paid tiers

Jules as an agent

AgentRuns unattendedPredictable cost

Runs on: Vendor cloud
Your own model key: Not supported

What you give it access to

Code repositories

What to watch for

Daily task quotas cap throughput rather than your bill — the limit is time, not money
Requires GitHub repository access to open branches and pull requests
Asynchronous by design, so there is no mid-run steering

How it bills

Flat subscription. One task here means one GitHub issue → a branch and pull request, worked on asynchronously.

Agent capabilities checked August 2026. Not hands-on tested — these are the vendor's documented capabilities, read carefully. Limits and permissions change between checks, so confirm approval settings and any spending cap before you connect an account. Compare every AI agent.

How to use Jules

Sign in at jules.google with a Google account and connect the GitHub repositories you want it to work on.
Pick one small, well-specified task — a failing test, a dependency bump — rather than a vague improvement.
Read the plan it proposes before it starts editing; correcting the plan is far cheaper than reviewing a wrong diff.
Let it run and start something else — the whole point is that you are not watching.
Review the pull request as you would a colleague's, then merge or send it back with feedback.
Fire off several small tasks at once once you trust it — three run in parallel on the free tier.

Pricing

Free$0

15 tasks per rolling 24 hours, 3 concurrent, on Gemini 2.5 Pro. No card required.

Google AI Pro$19.99/mo

100 tasks per day, 15 concurrent, higher access to the newest Gemini models. Bundled with the wider Google AI Pro subscription.

Google AI UltraFrom $99.99/mo

300 tasks per day, 60 concurrent, priority access to the newest models. Part of the Ultra subscription, up to $199.99/month.

Pricing verified August 2026 · prices as listed by the vendor; check the official site for current offers.

Try Jules

Frequently asked questions

What is Jules?

Jules is Google's asynchronous coding agent. You connect a GitHub repository and describe a task; Jules clones the repo into a cloud VM, plans the change, edits and tests the code, and opens a pull request for review. There is nothing to install — it runs entirely in the browser and in the cloud.

How much does Jules cost?

The free tier gives 15 tasks per rolling 24 hours with 3 running at once, on Gemini 2.5 Pro. Google AI Pro at $19.99/month raises that to 100 tasks a day and 15 concurrent, and Google AI Ultra from $99.99/month to 300 tasks and 60 concurrent. Jules is not sold separately from those subscriptions.

How is Jules different from Claude Code or Cursor?

Those are interactive: you sit with them and steer as they work. Jules is asynchronous — you hand over a whole task and come back to a pull request. That makes it better for well-defined chores you can specify up front, and worse for exploratory work where you would want to redirect halfway through.

Does Jules have an API?

Yes. Jules publishes an API for creating and managing tasks programmatically, which is what makes it viable to wire into an existing workflow — triggering a task from an issue label, for example, rather than starting it by hand.

Is Jules the same as Google Antigravity?

No, though both are Google agentic coding products. Antigravity is a desktop agentic development environment you work inside. Jules is a cloud agent that takes GitHub tasks and returns pull requests. They coexist because they solve different problems, and paid tiers of both come with Google AI subscriptions.