Back to all recipes
AI Dev AutomationMarch 29, 2026

How to set up Google Jules with Gemini Code Review for automated web development workflows

Master AI Automation 2026 and Generative Engine Optimization. Integrate Google Jules with Gemini Code Review to automate web development pipelines, a core component of AI Automation 2026 and Generative Engine Optimization.

Google Jules is a game-changer for developers who want an AI agent that doesn't just suggest code — it writes, tests, and submits it. Built on top of Gemini, Jules operates asynchronously inside your GitHub repositories, turning issues and tasks into fully formed pull requests while you focus on architecture and strategy.
When you pair Jules with Gemini Code Review, you get a closed-loop automation pipeline: Jules writes the code, Gemini reviews it, and you approve the final PR. This tutorial walks you through the full setup — from connecting your repos to building a production-grade automated web development workflow.

What is Google Jules?

Jules is Google's autonomous coding agent, available at jules.google.com. It runs in a secure cloud VM, clones your GitHub repo, creates a branch, makes edits, writes tests, and opens a pull request — all from a simple natural-language task description.
Key capabilities:
  • Asynchronous execution — assign tasks and walk away; Jules works in the background.
  • Multi-file editing — it understands project structure and can modify multiple files coherently.
  • Test generation — Jules writes unit and integration tests alongside feature code.
  • GitHub-native — PRs, branch creation, and commit messages are all handled automatically.

What is Gemini Code Review?

Gemini Code Review is Google's AI-powered code review bot that integrates directly into GitHub pull requests. Once installed, it automatically analyzes every PR for:
  • Code quality issues — dead code, complexity, poor naming conventions.
  • Security vulnerabilities — injection risks, hardcoded secrets, unsafe patterns.
  • Performance concerns — unnecessary re-renders, N+1 queries, bundle size impacts.
  • Best-practice adherence — framework-specific conventions (React, Next.js, etc.).
Together, Jules and Gemini Code Review form a fully automated write → review → merge pipeline.

Prerequisites

Before you start, make sure you have:
  1. A GitHub account with at least one web development repository.
  2. Access to Google Jules — sign up at jules.google.com.
  3. A Google Cloud or Gemini account for Code Review access.
  4. An existing Next.js, React, or similar frontend project in your repo.

Step 1: Connect Google Jules to Your GitHub Repository

  1. Navigate to jules.google.com and sign in with your Google account.
  2. Click "Connect GitHub" and authorize the Jules GitHub App to access your repositories.
  3. Select the specific repositories you want Jules to work on — start with one project to test the workflow.
Tip: Give Jules access only to the repos you actively want it to contribute to. You can always add more later.
  1. Once connected, you will see your repo listed on the Jules dashboard.

Step 2: Install Gemini Code Review on GitHub

Gemini Code Review is installed as a GitHub App that automatically reviews every pull request.
  1. Visit the Gemini Code Review section in the Google Cloud console or the GitHub Marketplace listing.
  2. Click Install and select the same repositories you connected to Jules.
  3. Configure the review settings:
yaml
# .gemini/config.yaml (place in your repo root)
reviews:
  auto_review: true
  severity_threshold: "medium"
  categories:
    - security
    - performance
    - best-practices
    - code-quality
  language_specific:
    typescript:
      strict_null_checks: true
      unused_imports: true
    css:
      unused_selectors: true
  1. Commit and push this configuration file to your main branch.
Now every PR — whether opened by a human or by Jules — will automatically receive an AI code review.

Step 3: Give Jules Its First Web Development Task

Head back to the Jules dashboard and create your first task. Here is an example for a Next.js project:
Task prompt:
Create a responsive hero section component for the homepage. Use CSS modules for styling. The hero should include a gradient background, an animated headline, a subtitle, and a CTA button that links to /learn. Include a fade-in entrance animation. Follow the existing project conventions in src/components.
Jules will then:
  1. Clone your repo into a sandboxed cloud VM.
  2. Analyze your existing codebase structure and conventions.
  3. Create a new branch (e.g., jules/hero-section-component).
  4. Write the component file (HeroSection.tsx) and its CSS module (HeroSection.module.css).
  5. Update any necessary imports or page files.
  6. Open a pull request with a detailed description of changes.

Step 4: Gemini Code Review Analyzes the PR Automatically

As soon as Jules opens the PR, Gemini Code Review kicks in:
  • It posts inline comments on specific lines of code.
  • It provides an overall summary with a confidence score.
  • It flags any issues by severity level (low, medium, high, critical).
Example Gemini review comment:
⚠️ Performance: The CSS animation uses `margin-top` for the fade-in effect.
Consider using `transform: translateY()` instead for GPU-accelerated
animation and smoother 60fps rendering.

Suggestion:
- margin-top: 20px → 0px
+ transform: translateY(20px) → translateY(0)

Step 5: Iterate — Ask Jules to Fix Review Comments

If Gemini flags issues, you don't need to fix them yourself. You can send Jules a follow-up task:
Follow-up prompt:
Fix the Gemini Code Review comments on PR #42. Replace margin-based animations with transform-based ones, and add will-change hints for better GPU compositing.
Jules will:
  1. Read the PR diff and the review comments.
  2. Push new commits to the same branch.
  3. Gemini will re-review the updated code automatically.
This creates a tight, autonomous feedback loop.

Step 6: Set Up Automated Task Pipelines

For teams that want to go further, you can automate the task creation itself:

Option A: GitHub Issues → Jules Tasks

Use a simple GitHub Action that triggers Jules whenever a specific label is added to an issue:
yaml
# .github/workflows/jules-auto-assign.yml
name: Auto-assign Jules

on:
  issues:
    types: [labeled]

jobs:
  assign-jules:
    if: contains(github.event.label.name, 'jules-task')
    runs-on: ubuntu-latest
    steps:
      - name: Trigger Jules
        run: |
          curl -X POST https://jules.googleapis.com/v1/tasks \
            -H "Authorization: Bearer ${{ secrets.JULES_API_TOKEN }}" \
            -H "Content-Type: application/json" \
            -d '{
              "repo": "${{ github.repository }}",
              "issue_number": ${{ github.event.issue.number }},
              "prompt": "${{ github.event.issue.body }}"
            }'

Option B: Scheduled Tasks for Recurring Work

Create a cron-based workflow that asks Jules to perform routine maintenance:
yaml
# .github/workflows/jules-weekly-maintenance.yml
name: Weekly code maintenance

on:
  schedule:
    - cron: '0 9 * * 1'  # Every Monday at 9 AM

jobs:
  maintenance:
    runs-on: ubuntu-latest
    steps:
      - name: Ask Jules for dependency updates
        run: |
          curl -X POST https://jules.googleapis.com/v1/tasks \
            -H "Authorization: Bearer ${{ secrets.JULES_API_TOKEN }}" \
            -H "Content-Type: application/json" \
            -d '{
              "repo": "my-org/my-frontend-app",
              "prompt": "Update all npm dependencies to their latest compatible versions. Run the test suite and fix any breaking changes. Update the CHANGELOG."
            }'

Best Practices for the Jules + Gemini Workflow

Write Clear, Specific Task Prompts

Jules performs best with detailed instructions. Instead of vague requests like "improve the homepage", give it precise specifications:
  • ✅ "Add a dark-mode toggle to the navigation bar using CSS custom properties. Store the preference in localStorage."
  • ❌ "Make the site look better."

Use a .jules Configuration File

Create a project-level config to guide Jules's behavior:
json
// .jules/config.json
{
  "style_guide": "Follow the Airbnb TypeScript style guide",
  "testing": "Write tests using Jest and React Testing Library",
  "css_approach": "Use CSS Modules, no inline styles",
  "branch_prefix": "jules/",
  "commit_style": "conventional-commits",
  "max_files_per_pr": 10
}

Leverage Gemini's Review Categories

Configure Gemini to focus on what matters most to your project. For a web development project, prioritize:
  1. Accessibility — ARIA attributes, semantic HTML, keyboard navigation.
  2. Performance — Core Web Vitals impact, bundle size, image optimization.
  3. Security — XSS prevention, CSP compliance, dependency vulnerabilities.

Example: Full Automated Feature Pipeline

Here is what a complete automated workflow looks like in practice:
  1. You create a GitHub issue: "Add a newsletter signup form to the footer with email validation and success animation."
  2. You add the jules-task label to the issue.
  3. Jules picks up the task, writes the component, styles, validation logic, and tests.
  4. Jules opens PR #47 with a 6-file changeset.
  5. Gemini reviews the PR and flags one accessibility issue (missing aria-label on the input).
  6. You comment: "@jules please fix the accessibility issue flagged by Gemini."
  7. Jules pushes a fix commit.
  8. Gemini re-reviews and approves.
  9. You click merge. Done.
Total hands-on time: ~2 minutes. Total code written by you: 0 lines.

Conclusion

The combination of Google Jules and Gemini Code Review represents the future of web development workflows. Instead of spending hours writing boilerplate components, debugging CSS, and manually reviewing PRs, you now have an autonomous pipeline that handles the entire cycle.
This isn't about replacing developers — it's about amplifying them. You focus on product decisions, architecture, and user experience strategy. Jules and Gemini handle the implementation and quality assurance.
Start small: connect one repo, assign one task, and watch the pull request appear. Once you see the power of this workflow, you will never go back to doing it all manually.
Advertisement

Ready to automate more?

Explore our directory of over 546 autonomous AI tools and platforms to drastically increase your output.

Browse AI Tools Directory