Debugging with AI

Working Code

First, let's create code with an intentional bug:

> Create a bug-demo.js file with this code:
> const users = [{ name: "Alice", age: 30 }, { name: "Bob", age: 25 }];
> console.log(users.find(u => u.name === "Charlie").age);

Now run it:

> Run node bug-demo.js

A TypeError: Cannot read properties of undefined error occurs. Claude sees the error and immediately analyzes the cause:

> Fix this error

Claude identifies that find can return undefined and applies optional chaining or a null check.

Try It Yourself

Let's try a more realistic scenario. Assume an error occurs in a real project:

> Run npm run build

If there's a build error, Claude reads the error log and analyzes the cause. If your project has no errors, create one:

> Intentionally break an import path in server.js, then run it to see the error.

After seeing the error:

> Fix this error

"Why?" — Why Claude Code Excels at Debugging

Claude Code doesn't just look at the error message. Because it knows the full project context:

1. It checks related files, not just the file where the error occurred.
2. It traces import relationships to find the root cause.
3. It detects mismatches with configuration files (tsconfig, package.json, etc.).

Effective Debugging Request Patterns

> Run npm test and fix the failing tests

> Analyze this error log: [paste error message]

> The users API returns a 500 error — find the cause

> Add console.log statements to this function to trace the data flow

Debugging Workflow

Error occurs → Ask Claude to run → Automatic error analysis → Fix suggestion → Approve → Re-run to verify

For complex bugs that aren't resolved in one pass, take a step-by-step approach:

> Analyze the cause of this error. Don't fix it yet.

Once you understand the cause:

> OK, fix it using the ~ approach

Deep Dive