Parallel Tasks
OpenClaw supports executing multiple sub-tasks in parallel within a single conversation, dramatically improving complex task throughput.
How It Works
When the agent receives a complex task, it breaks it into independent sub-tasks. Those without dependencies run concurrently:
User Request → Agent Analysis
├── Sub-task A (file read) ─┐
├── Sub-task B (API call) ─┼── Parallel
└── Sub-task C (web scrape) ─┘
→ Merge → Return
Configuration
{
agents: {
defaults: {
tools: {
parallel_execution: true,
max_concurrent: 5,
},
},
},
}
Best Scenarios for Parallelism
Multi-file processing: Analyze multiple files simultaneously
User: Review all Python files in src/ for code quality
Multiple API calls: Query several data sources at once
User: Check weather in Beijing, Shanghai, and Guangzhou
Batch web scraping: Open multiple pages concurrently
User: Scrape titles from these 5 competitor websites
Task Dependencies
The agent automatically detects dependencies and sequences them:
Step 1: Read config (must complete first)
Step 2: Parallel execution
├── Call API A with config
├── Call API B with config
└── Query database with config
Step 3: Aggregate results into report (waits for Step 2)
Combining with Cron
Pair parallel tasks with scheduled execution:
openclaw cron add --schedule "0 9 * * *" \
--message "Check all monitors in parallel, generate daily report" \
--channel telegram --target <ChatID>