Course Home / Module 1: Fundamentals

1.5: Parallel Agents

25 min
Interactive: start 1-5

What Are Agents?

Agents are independent instances of Claude running in parallel. Instead of processing tasks one at a time, you can run many simultaneously—dramatically speeding up batch work.

Example: Process 10 tasks in 5 minutes instead of 50 minutes.

When to Use Agents

Good use cases

  • Many similar files that need the same operation
  • Parallel research across multiple sources
  • Batch document generation (reports, summaries, analyses)
  • Multi-source analysis where sources don’t depend on each other

When NOT to use agents

  • Single tasks (no benefit from parallelization)
  • Sequential work where each step depends on the previous
  • Quick operations that finish in seconds anyway

Five Orchestration Patterns

1. Fan Out

Apply identical tasks to many files simultaneously.

For each PDF in /contracts, extract key terms and save to /summaries

2. Specialized Roles

Different agents analyze different aspects of the same data.

Have one agent analyze financial data, another analyze legal risks,
and a third analyze market positioning from @company-report.pdf

3. Parallel Research

Research multiple entities at the same time.

Research these five competitors simultaneously and create a summary for each:
- Company A
- Company B
- Company C
- Company D
- Company E

4. Batch Generation

Create multiple similar documents at once.

Generate personalized outreach emails for each contact in @leads.csv

5. Validation Pipeline

Review multiple documents against the same criteria.

Check each proposal in /submissions against our requirements in @criteria.md

Critical: Always Synthesize

After all agents complete, always synthesize results.

Raw outputs from parallel agents need combining to reveal patterns:

After all agents complete, synthesize results into @summary.md

Without synthesis, you just have many separate outputs—not insights.

Practical Examples

Job Analysis

Analyze the 20 most recent job postings for "Product Manager" and identify:
- Common required skills
- Salary ranges
- Company size patterns

Synthesize findings into @job-market-analysis.md

Vendor Evaluation

For each vendor in @vendors.csv:
- Research their pricing
- Find customer reviews
- Check their compliance certifications

Create a comparison table in @vendor-comparison.md

Article Summarization

Summarize each article in /research-papers and identify common themes across all of them

How Agents Work Technically

When you request parallel processing:

  1. Claude spawns multiple independent instances
  2. Each instance works on its assigned task
  3. Results are collected when all complete
  4. A synthesis step combines the outputs

You don’t need to manage this—just describe what you want done.

Performance Tips

  • Group similar tasks for maximum efficiency
  • Set clear output formats so synthesis is easier
  • Include the synthesis step in your original request
  • Monitor for failures—if one agent fails, others may still complete

Limitations

  • Agents can’t communicate with each other during processing
  • Very large batches may hit rate limits
  • Complex interdependent tasks should be sequential, not parallel

Ready to try parallel processing? Run start 1-5 for hands-on practice with agents.