AI

Claude Workflow: Researching a Feature

An example workflow showing how Claude uses codebase exploration and planning agents to research, design, and propose a new feature implementation plan.

How It Works

What you do: Ask Claude a question such as "How should we implement real-time notifications?"

What happens:

  1. Claude reads your CLAUDE.md instructions and learns about your tech stack, existing patterns, and architectural constraints.
  2. An Explore agent is dispatched to search your codebase for:
    • Existing notification systems
    • WebSocket or server-sent event implementations
    • Message queue integrations
  3. A Plan agent uses the information it has gathered to design an implementation approach that fits your existing patterns.
  4. The agent may use the Sonnet model, which offers a good balance of capability and speed for this kind of research task.
  5. It presents a plan showing which files need changes, what patterns to follow (based on your instructions), and the key design decisions.
  6. You approve the plan, and the agent proceeds with implementation.
← Older
Claude Workflow: Reviewing a Pull Request
Newer →
Claude Workflow: Add a New Page

Newsletter

A weekly newsletter on React, Next.js, AI-assisted development, and engineering. No spam, unsubscribe any time.