Jules

You manage and orchestrate multiple Jules AI instances with a TypeScript-based MCP server that handles code generation, bug fixing, and reviews. This server connects to Jules, coordinates workers, tracks activity, and provides safe, validated interactions across your development tasks.

How to use

You will use an MCP client to interact with the Jules MCP server. Start by running the MCP server locally or connect to a remote MCP endpoint, then create workers, send messages between them, generate code, fix bugs, and review results. All actions go through a consistent tool interface that validates inputs and tracks progress in real time.

How to install

# Prerequisites
node -v  # ensure Node.js 18+ is installed
npm -v

# Quick start: run directly with npx (recommended for testing)
npx jules-mcp

# Or install globally and run
npm install -g jules-mcp
jules-mcp

# Local development workflow (example)
git clone https://github.com/access_aipro/jules-mcp-npx
cd jules-mcp
npm install
export JULES_API_KEY="your-api-key-here"
npm run dev
npm run build
npm start

Configuration

Set environment variables to configure how the MCP server authenticates with Jules, controls limits, and tunes behavior. These values are read at startup and govern API access, logging, rate limiting, and cost tracking.

# Required
JULES_API_KEY=your-api-key-here

# Optional
SERVICE_PORT=8085
LOG_LEVEL=INFO
JULES_API_BASE_URL=https://jules.googleapis.com
MAX_COST_PER_HOUR=10.00
DAILY_COST_LIMIT=100.00
CACHE_TTL=3600
RATE_LIMIT_REQUESTS=60
CODE_VALIDATION_ENABLED=true
COST_TRACKING_ENABLED=true

Examples and usage patterns

Use the MCP client to perform common workflows such as creating Jules workers, generating code, fixing bugs, or reviewing code. The following patterns illustrate typical interactions.

# Basic code generation
result = await jules_mcp.call_tool("jules_generate_code", {
  "prompt": "Create a React component for user login form with TypeScript",
  "language": "typescript",
  "context": {
    "framework": "react",
    "styling": "tailwind",
    "validation": "yup"
  }
})

# Bug fixing
result = await jules_mcp.call_tool("jules_fix_bug", {
  "code": "existing-code-with-leak.js",
  "error_description": "Memory usage increases over time",
  "expected_behavior": "Constant memory usage"
})

# Code review
result = await jules_mcp.call_tool("jules_review_code", {
  "code": "api-endpoint.py",
  "language": "python",
  "focus_areas": ["security", "performance", "error_handling"]
})

Security, testing, and monitoring notes

Monitor health, track costs, and ensure secure interactions between workers. Use health checks to verify readiness, view active workers, and inspect cost tracking to stay within budget.

GET /health
GET /health/detailed
GET /health/cost-tracker

Troubleshooting

If you encounter startup or connection issues, verify that the required environment variables are set, the API key is valid, and the MCP server process has the necessary permissions to access Jules services. Check logs for validation errors, worker failures, or rate limit events, and adjust configuration values accordingly.

Available tools

jules_create_worker

Create Jules AI workers for specific implementation tasks, with roles such as Maestro, Crew, and Evaluator

jules_send_direct_message

Send direct messages between workers to coordinate actions

jules_estimate_work

Create an Evaluator worker to estimate task effort and scope

jules_store_memory

Store shared memory values accessible across worker sessions

jules_read_memory

Read shared memory values from the shared store

jules_create_branch

Create a new Git branch to support staged orchestration workflows

jules_merge_branch

Merge a feature branch into a target branch to integrate changes

jules_list_branches

List all local Git branches in the workspace

jules_generate_code

Generate code tailored to specific requirements with context awareness

jules_fix_bug

Analyze and fix bugs in existing codebases

jules_review_code

Perform comprehensive code reviews focusing on security, quality, and maintainability

jules_get_status

Check worker statuses and progress across all active workers