Skip to main content
Core Concept

Parallel Agents

Run multiple agents simultaneously for dramatically faster code exploration and analysis.

Sequential Exploration (Slow)

→ Search auth code

[wait 30 seconds]

→ Check error logs

[wait 30 seconds]

→ Review recent changes

[wait 30 seconds]

Total: ~90 seconds

Parallel Agents (Fast)

→ Launch 3 agents at once

Agent 1: Search auth code

Agent 2: Check error logs

Agent 3: Review recent changes

[all run simultaneously]

Total: ~30 seconds (3x faster!)

When to Use Parallel Agents

🔍

Code Exploration

Search multiple parts of the codebase simultaneously—auth, database, API, frontend.

🐛

Bug Investigation

One agent checks error logs, another reviews recent commits, another traces the code path.

📊

Code Review

Different agents review security, performance, and test coverage in parallel.

📝

Documentation

Generate README, API docs, and usage examples all at once.

How It Works

1

Claude Identifies Independent Tasks

Tasks that don't depend on each other can run in parallel

2

Agents Fork into Isolated Contexts

Each agent works independently without affecting your main conversation

3

Results Are Combined

Claude synthesizes findings from all agents into a cohesive response

The Key Benefit

max(individual times) ≠ sum(all times)

Parallel agents complete in the time of the slowest task, not the sum of all tasks. This is the power of concurrent execution.

Learn More

Discover advanced patterns for parallel agent workflows and when to use them.