Parallel Exploring

Parallel Explore

Use this skill when you need to understand a large or unfamiliar codebase quickly — onboarding onto a new project, investigating how a feature works across layers, or mapping the architecture.

How It Works

Cursor's explore subagent is a fast, read-only agent optimized for searching and reading code. You can launch multiple explore agents in a single message and they run concurrently, each investigating a different area.

Steps

1. Identify the areas to explore — break the codebase into logical zones. For a typical full-stack app:

   • Frontend: components, pages, routing, state management
   • Backend: API routes, database models, middleware, auth
   • Infrastructure: CI/CD, Docker, deployment config
   • Shared: types, utilities, constants

2. Launch parallel explore agents — use the Task tool with subagent_type: "explore" for each area. Launch them all in one message:

   Task 1: "Explore the frontend — find the main pages, routing setup, state management approach,
            and UI component library. Check src/app/, src/components/, src/pages/. Report the
            framework, router, styling approach, and key components."
   
   Task 2: "Explore the backend — find the API routes, database setup, ORM, auth middleware,
            and data models. Check src/server/, src/api/, lib/, prisma/. Report the framework,
            database, auth strategy, and key endpoints."
   
   Task 3: "Explore the infrastructure — find CI/CD config, Docker setup, deployment targets,
            and environment variable management. Check .github/, docker*, *.config.*, .env*.
            Report the deploy target, CI provider, and any IaC."
   

3. Synthesize the results — when all agents return, combine their findings into a coherent picture:

   • Tech stack summary (frontend, backend, database, infra)
   • Architecture diagram (describe the data flow)
   • Key files and entry points
   • Potential concerns or tech debt

Other Use Cases

• Cross-cutting investigation: "Where is user authentication checked?" — launch agents to search the frontend (route guards), backend (middleware), and database (session storage) simultaneously.
• Dependency audit: launch agents to check different parts of the dependency tree for outdated packages, security issues, and unused imports.
• Migration planning: have agents simultaneously assess the frontend, backend, and tests to estimate the scope of a framework migration.

Notes

• Explore agents are read-only — they can't modify files.
• Use thoroughness: "very thorough" in the prompt for comprehensive analysis.
• Each agent has its own context window, so they can each read many files without running out of space.
• For a single focused question, just use Grep or SemanticSearch directly — subagents are for broad exploration.