Skip to main content
Glama
README.md
# Google Jules MCP

A Model Context Protocol (MCP) server for automating Google Jules - the AI coding assistant. This MCP enables seamless integration with Jules for task creation, code review automation, repository management, and AI-powered development workflows.

## Features

### šŸŽÆ **Task Management**
- **Create Tasks**: Automatically create Jules tasks with repository and description
- **Monitor Progress**: Track task status and get real-time updates
- **Approve Plans**: Review and approve Jules execution plans
- **Resume Tasks**: Resume paused or interrupted tasks
- **Bulk Operations**: Create multiple tasks efficiently

### šŸ”§ **Code Operations**
- **Code Analysis**: Analyze code changes and diffs
- **Branch Management**: Handle repository branches and configurations
- **Source Navigation**: Browse and analyze source files
- **Review Automation**: Automate code review workflows

### šŸ’¬ **Interactive Communication**
- **Send Messages**: Send instructions and feedback to Jules
- **Chat History**: Track conversation history with Jules
- **Context Extraction**: Extract relevant context from task discussions

### šŸ“Š **Project Management**
- **Task Listing**: List and filter tasks by status
- **Progress Tracking**: Monitor development progress across projects
- **Data Persistence**: Local storage of task data and history

### šŸ” **Session Management** (NEW!)
- **Multiple Session Modes**: Fresh, Chrome profile, cookies, persistent, and Browserbase
- **Google Authentication**: Seamless login with existing Google sessions
- **Cookie Management**: Extract, save, and restore authentication cookies
- **Remote Browser Support**: Use Browserbase for cloud deployments
- **Cross-Platform**: Works locally and in cloud environments

## Available Tools

| Tool | Description |
|------|-------------|
| **Task Management** ||
| `jules_create_task` | Create a new Jules task with repository and description |
| `jules_get_task` | Get detailed information about a specific task |
| `jules_send_message` | Send messages/instructions to Jules in active tasks |
| `jules_approve_plan` | Approve Jules execution plans |
| `jules_resume_task` | Resume paused tasks |
| `jules_list_tasks` | List tasks with filtering options |
| `jules_analyze_code` | Analyze code changes and project structure |
| `jules_bulk_create_tasks` | Create multiple tasks from a list |
| **Session & Authentication** ||
| `jules_get_cookies` | Get current browser cookies for session persistence |
| `jules_set_cookies` | Set browser cookies from string/JSON for authentication |
| `jules_session_info` | Get current session configuration and status |
| **Debugging** ||
| `jules_screenshot` | Take debugging screenshots |

## Installation

### Prerequisites
- Node.js 18+ 
- TypeScript
- Git access to repositories you want to manage

### Setup

```bash
# Clone the repository
git clone https://github.com/samihalawa/google-jules-mcp.git
cd google-jules-mcp

# Install dependencies
npm install

# Build the project
npm run build

# Test the installation
npm test
```

## šŸ” Session Management & Authentication

### Session Modes

The MCP supports 5 different session management modes:

| Mode | Description | Use Case |
|------|-------------|----------|
| `fresh` | New browser session each time | Testing, no authentication needed |
| `chrome-profile` | Use existing Chrome profile | Local development with existing Google login |
| `cookies` | Save/load cookies to file | Persistent authentication without full profile |
| `persistent` | Save browser data to directory | Local development with full session persistence |
| `browserbase` | Remote browser session | Cloud deployments, Smithery hosting |

### Configuration Options

#### 🌐 **Browserbase (Recommended for Smithery)**

Perfect for remote deployments and cloud hosting:

```bash
SESSION_MODE=browserbase
BROWSERBASE_API_KEY=bb_live_g3i-b4WPFh__E3cErKE5rO-jWds  # Your API key
BROWSERBASE_PROJECT_ID=d718e85f-be7b-497d-9123-b1bbf798f1bb  # Your project ID
BROWSERBASE_SESSION_ID=                    # Optional: use existing session
```

#### šŸŖ **Cookie Authentication (Best for Manual Setup)**

Extract cookies from your browser and set them as environment variable:

```bash
SESSION_MODE=cookies
GOOGLE_AUTH_COOKIES="session_id=abc123; domain=.google.com; auth_token=xyz789; domain=.google.com"
COOKIES_PATH=~/.jules-mcp/cookies.json     # File to save/load cookies
```

#### šŸŒ **Chrome Profile (Local Development)**

Use your existing Chrome profile:

```bash
SESSION_MODE=chrome-profile
CHROME_USER_DATA_DIR=/Users/yourname/Library/Application Support/Google/Chrome/Default
```

#### šŸ’¾ **Persistent Browser Data**

Save browser data to a specific directory:

```bash
SESSION_MODE=persistent
CHROME_USER_DATA_DIR=~/.jules-mcp/browser-data  # Custom browser data directory
```

### How to Get Google Authentication Cookies

1. **Log in to Jules**: Visit https://jules.google.com and log in
2. **Open Developer Tools**: Press F12 or Cmd+Option+I
3. **Go to Application/Storage tab**
4. **Find Cookies**: Look for `.google.com` cookies
5. **Copy Important Cookies**: Look for cookies like:
   - `session_id` or `sessionid`
   - `auth_token` or `authuser`
   - `SID`, `HSID`, `SSID`
   - `SAPISID`, `APISID`

**Format for environment variable:**
```bash
GOOGLE_AUTH_COOKIES="cookie1=value1; domain=.google.com; cookie2=value2; domain=.google.com"
```

### Environment Configuration

Create a `.env` file or set environment variables:

```bash
# Browser Configuration
HEADLESS=true              # Run browser in headless mode
TIMEOUT=30000              # Browser timeout in milliseconds
DEBUG=false                # Enable debug mode with screenshots

# Session Management
SESSION_MODE=browserbase   # fresh | chrome-profile | cookies | persistent | browserbase

# Browserbase Configuration (for remote/cloud deployments)
BROWSERBASE_API_KEY=your_api_key
BROWSERBASE_PROJECT_ID=your_project_id
BROWSERBASE_SESSION_ID=optional_existing_session

# Cookie Authentication
GOOGLE_AUTH_COOKIES="session_id=abc; domain=.google.com"
COOKIES_PATH=~/.jules-mcp/cookies.json

# Chrome Profile (local development)
CHROME_USER_DATA_DIR=/path/to/chrome/profile

# Data Storage
JULES_DATA_PATH=~/.jules-mcp/data.json  # Custom data storage path
```

## Usage Examples

### 1. Create a New Task

```javascript
// Create a task to fix a bug
{
  "name": "jules_create_task",
  "arguments": {
    "description": "Fix the login authentication bug in the user dashboard. The issue occurs when users try to log in with special characters in their password.",
    "repository": "mycompany/webapp",
    "branch": "main"
  }
}
```

### 2. Monitor Task Progress

```javascript
// Get task details and progress
{
  "name": "jules_get_task",
  "arguments": {
    "taskId": "9103172019911831130"
  }
}
```

### 3. Send Instructions to Jules

```javascript
// Send additional context or instructions
{
  "name": "jules_send_message",
  "arguments": {
    "taskId": "9103172019911831130",
    "message": "Please also add unit tests for the authentication fix and ensure backward compatibility."
  }
}
```

### 4. Bulk Task Creation

```javascript
// Create multiple tasks at once
{
  "name": "jules_bulk_create_tasks",
  "arguments": {
    "tasks": [
      {
        "description": "Add dark mode support to the UI",
        "repository": "mycompany/frontend",
        "branch": "feature/dark-mode"
      },
      {
        "description": "Optimize database queries for user search",
        "repository": "mycompany/backend",
        "branch": "performance/search"
      }
    ]
  }
}
```

### 5. List and Filter Tasks

```javascript
// List active tasks
{
  "name": "jules_list_tasks",
  "arguments": {
    "status": "in_progress",
    "limit": 10
  }
}
```

### 6. Session Management Examples

#### Check Session Status
```javascript
{
  "name": "jules_session_info",
  "arguments": {}
}
```

#### Get Current Cookies (for backup)
```javascript
{
  "name": "jules_get_cookies",
  "arguments": {
    "format": "string"  // or "json"
  }
}
```

#### Set Authentication Cookies
```javascript
{
  "name": "jules_set_cookies",
  "arguments": {
    "cookies": "session_id=abc123; domain=.google.com; auth_token=xyz789; domain=.google.com",
    "format": "string"
  }
}
```

## MCP Resources

The server provides useful resources for context:

- `jules://schemas/task` - Complete task data model
- `jules://current/active-tasks` - Live list of active tasks
- `jules://templates/common-tasks` - Template examples for common development tasks

## Common Task Templates

The MCP includes templates for common development scenarios:

- **Bug Fix**: `"Fix the [specific issue] in [filename]. The problem is [description]."`
- **Feature Add**: `"Add [feature name] functionality to [location]. Requirements: [list requirements]."`
- **Refactor**: `"Refactor [component/function] to improve [performance/readability/maintainability]."`
- **Testing**: `"Add comprehensive tests for [component/function] covering [test cases]."`
- **Documentation**: `"Update documentation for [component] to include [new features/changes]."`

## Integration with Claude Code

### Local Integration

```json
{
  "mcpServers": {
    "google-jules-mcp": {
      "command": "node",
      "args": ["path/to/google-jules-mcp/dist/index.js"],
      "env": {
        "HEADLESS": "true",
        "SESSION_MODE": "cookies",
        "GOOGLE_AUTH_COOKIES": "your_cookies_here",
        "DEBUG": "false"
      }
    }
  }
}
```

## 🌐 Smithery Deployment

### Deploy to Smithery.ai

The MCP is fully configured for Smithery deployment with comprehensive session management:

1. **Fork/Clone** this repository
2. **Deploy to Smithery**: Visit [smithery.ai](https://smithery.ai) and connect your repo
3. **Configure Session Management** in Smithery settings:

#### Option A: Browserbase (Recommended)
```bash
SESSION_MODE=browserbase
BROWSERBASE_API_KEY=bb_live_g3i-b4WPFh__E3cErKE5rO-jWds
BROWSERBASE_PROJECT_ID=d718e85f-be7b-497d-9123-b1bbf798f1bb
```

#### Option B: Cookie Authentication
```bash
SESSION_MODE=cookies
GOOGLE_AUTH_COOKIES="session_id=abc123; domain=.google.com; auth_token=xyz789; domain=.google.com"
```

4. **Access Remotely**: Use your deployed MCP from any Claude Code instance

### Benefits of Smithery + Browserbase

- āœ… **No Local Browser**: Runs entirely in the cloud
- āœ… **Persistent Sessions**: Maintain Google authentication across deployments
- āœ… **Global Access**: Use from anywhere with internet connection
- āœ… **Auto-scaling**: Handles multiple concurrent requests
- āœ… **Zero Setup**: No local dependencies or configuration needed

## Troubleshooting

### Common Issues

1. **Browser Automation Fails**
   - Ensure you have proper access to `jules.google.com`
   - Check if you're logged into your Google account
   - Try running with `HEADLESS=false` to see what's happening

2. **Task Creation Fails**
   - Verify repository names are correct (`owner/repo-name` format)
   - Ensure you have access to the specified repositories
   - Check that branches exist

3. **Permission Errors**
   - Make sure you have write access to the data storage path
   - Verify repository permissions in GitHub

### Debug Mode

Enable debug mode for troubleshooting:

```bash
DEBUG=true HEADLESS=false npm start
```

This will:
- Show browser interactions visually
- Take screenshots on errors
- Provide detailed logging

## Development

### Project Structure

```
google-jules-mcp/
ā”œā”€ā”€ src/
│   └── index.ts          # Main MCP server implementation
ā”œā”€ā”€ docs/
│   └── referencerecordings/  # Browser automation references
ā”œā”€ā”€ scripts/
│   └── test-mcp.js       # Testing script
ā”œā”€ā”€ dist/                 # Compiled output
ā”œā”€ā”€ package.json
ā”œā”€ā”€ tsconfig.json
└── smithery.yaml         # MCP deployment config
```

### Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

### Testing

```bash
# Run the test suite
npm test

# Build and test
npm run build && npm test

# Development mode with file watching
npm run dev
```

## Architecture

The MCP follows established patterns:

- **Browser Automation**: Uses Playwright for reliable web automation
- **Data Persistence**: Local JSON storage for task tracking
- **Error Handling**: Comprehensive error handling with meaningful messages
- **Resource Management**: Proper browser lifecycle management
- **Security**: No credential storage, relies on browser session

## Workflow Integration

This MCP is designed to integrate with development workflows:

1. **Issue Tracking → Jules Tasks**: Convert GitHub issues to Jules tasks
2. **Code Review → Automation**: Automate code review processes
3. **CI/CD Integration**: Trigger Jules tasks from deployment pipelines
4. **Team Collaboration**: Share Jules task management across teams

## License

MIT License - see LICENSE file for details.

## Acknowledgments

- Built with the [Model Context Protocol SDK](https://github.com/modelcontextprotocol/sdk)
- Inspired by the tusclasesparticulares-mcp implementation patterns
- Browser automation powered by [Playwright](https://playwright.dev/)

---

**Note**: This MCP requires access to Google Jules. Ensure you have appropriate permissions and access to the repositories you want to manage.

TDQS

A3.6/5.0

Scored across 13 tools

Disambiguation5/5

Every tool has a clearly distinct purpose with no ambiguity. Tools like jules_analyze_code, jules_approve_plan, jules_create_task, and jules_get_task each target specific actions on different aspects of the Jules system, and there is no overlap in functionality that would cause confusion.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern with the prefix 'jules_' and snake_case throughout. Examples include jules_create_task, jules_list_tasks, and jules_send_message, making the naming predictable and easy to understand.

Tool Count5/5

With 13 tools, the set is well-scoped for managing Google Jules tasks, covering creation, retrieval, control, and session management. Each tool earns its place without feeling excessive or insufficient for the domain.

Completeness5/5

The tool surface provides complete coverage for the Jules task lifecycle, including create, get, list, approve, resume, and send messages, along with session setup and authentication tools. There are no obvious gaps that would hinder agent workflows.