DEV Community

Cover image for Tried IBM Bob, an AI agentic coding assistant. Built a basic Java microservices architecture in ~15 minutes
Abhay Rao
Abhay Rao

Posted on

Tried IBM Bob, an AI agentic coding assistant. Built a basic Java microservices architecture in ~15 minutes

#ai #programming #java #ibmbob

From a Few Prompts to a Running Java Microservices App with IBM Bob

I've been trying out IBM Bob, an AI coding assistant, and recently watched a demo that caught my attention.

A lot of coding demos focus on generating a single function or API. This one was different. The goal was to start with a high-level idea and see how far a few prompts could take a Spring Boot microservices project.

Step 1: Start with the Architecture

The first prompt was pretty simple:

Design a customer order management system with separate services for customers, orders, inventory, and notifications.
Enter fullscreen mode Exit fullscreen mode

Bob broke the problem down into multiple services and proposed a reasonable microservices architecture.

It felt similar to the kind of architecture sketch you'd normally put together before writing any code.

Step 2: Generate the Services

The next step was expanding that architecture into actual services.

Generate Spring Boot microservices for the architecture above.
Include REST APIs, service layers, JPA repositories,
Swagger documentation, and standardized error handling.
Enter fullscreen mode Exit fullscreen mode

From there, Bob generated:

  • REST controllers
  • Service layer code
  • JPA repositories
  • OpenAPI / Swagger documentation
  • Error handling patterns

What surprised me wasn't any individual component. It was how much of the repetitive setup work was handled automatically.

Step 3: Add Resilience

Once the services were in place, the focus shifted to reliability.

Add retry patterns and health checks to the order service.
Enter fullscreen mode Exit fullscreen mode

Bob then generated:

  • Resilience4j retry configuration
  • Custom health indicators
  • Metrics endpoints
  • Additional observability configuration

These are the kinds of things that often get postponed until later in a project, so it was interesting to see them added through a follow-up prompt.

Step 4: Run and Validate

Finally, the generated services were started and validated.

Start all services, verify health endpoints,
and open the generated API documentation.
Enter fullscreen mode Exit fullscreen mode

The workflow included:

  • Starting the services
  • Checking health endpoints
  • Reviewing API documentation
  • Verifying service communication

At that point, there was a runnable system rather than just generated source files.

What Stood Out

The interesting part wasn't that AI generated Java code.

We've all seen that before.

What stood out was the progression:

Idea
  ↓
Architecture
  ↓
Multiple Services
  ↓
Resilience
  ↓
Observability
  ↓
Runnable System
Enter fullscreen mode Exit fullscreen mode

All of that came from a small series of prompts rather than manually building everything piece by piece.

Is it production-ready? Obviously not.

There are still questions around security, testing, deployment, performance, governance, and long-term maintainability.

But as a way to bootstrap a cloud-native Java project, it was impressive to see how much groundwork could be laid in a short time.

Curious to Hear From Others

Has anyone here used IBM Bob, Claude Code, Cursor, GitHub Copilot, or similar agentic coding tools for Java/Spring Boot projects?

What's the most useful thing you've had them build, and where do you still find yourself taking over manually?

Top comments (0)