Leveraging Sub-Agents for Complex Tasks

Sub-agents are a powerful feature in OpenClaw that allow you to delegate complex or time-consuming tasks without blocking your main conversation. Think of them as background workers that handle tasks independently while you continue working in your primary session.

What Are Sub-Agents?

Sub-agents are isolated sessions spawned from your main agent to perform specific tasks. When you spawn a sub-agent, it runs in the background, completes its assigned work, and reports back the results to your main conversation. This allows for parallel processing and efficient task management.

Key characteristics:

• Isolated execution: Each sub-agent runs in its own session with separate context
• Non-blocking: Your main conversation continues uninterrupted
• Result announcement: Completed tasks are summarized and delivered back to your chat
• Auto-archived: Sessions are cleaned up automatically after completion

Common Use Cases

1. Research and Information Gathering

When you need to research a topic while continuing your current conversation:

"Spawn a sub-agent to research the latest Node.js release notes and summarize the key changes"

The sub-agent will work on the research while you continue discussing other topics. Once complete, it will return a concise summary.

2. File Processing and Analysis

For analyzing multiple files or processing large datasets:

"Spawn a sub-agent to analyze all log files from today and identify any error patterns"

This is particularly useful for system administrators or developers who need to process server logs, codebases, or data files.

3. Multi-Step Operations

Complex workflows that involve multiple steps can be delegated to sub-agents:

"Spawn a sub-agent to update all documentation, run tests, and generate a deployment report"

4. Parallel Task Execution

When you have several independent tasks, you can spawn multiple sub-agents to work on them simultaneously:

"Spawn sub-agents to check server status, monitor application performance, and review security logs"

How to Use Sub-Agents

Basic Spawning

Simply ask your agent to spawn a sub-agent with a clear task description:

"Please spawn a sub-agent to [task description]"

For example:

"Spawn a sub-agent to compile a list of the top 10 AI tools for developers and their key features"

Advanced Options

You can specify additional parameters for more control:

• Model selection: "Use gpt-5.2 for this sub-agent" - specifies which model to use
• Timeout: "Set a 5-minute timeout" - prevents tasks from running indefinitely
• Thinking level: "Use high thinking mode" - for more complex reasoning tasks

Example with options:

"Spawn a sub-agent to analyze this codebase for security vulnerabilities. Use claude-opus-4-6 and set a 10-minute timeout."

Benefits of Using Sub-Agents

1. Increased productivity: Handle multiple tasks simultaneously
2. Better resource management: Offload intensive tasks from your main session
3. Cleaner workflow: Keep your primary conversation focused while background tasks complete
4. Cost efficiency: Use less expensive models for sub-agents when appropriate

Best Practices

• Be specific with task descriptions: Clear instructions lead to better results
• Monitor active sub-agents: Use /subagents list to see what's running
• Set appropriate timeouts: Prevent runaway processes
• Use the right model: Match the model capability to the task complexity
• Clean up when needed: Use /subagents stop <id> to terminate tasks if necessary

Managing Your Sub-Agents

OpenClaw provides several commands to manage your sub-agents:

• /subagents list - View all active and completed sub-agents
• /subagents stop <id> - Stop a running sub-agent
• /subagents log <id> - View the transcript of a sub-agent
• /subagents info <id> - Get detailed information about a sub-agent

Real-World Example

Imagine you're preparing for a team meeting and need several pieces of information:

1. Current project status
2. Recent customer feedback
3. Server performance metrics
4. Upcoming deadlines

Instead of handling these sequentially, you could spawn four sub-agents:

"Spawn sub-agents to: (1) get the latest project status from our task board, (2) compile recent customer feedback from support tickets, (3) retrieve current server performance metrics, and (4) list all upcoming deadlines. Report back when complete."

While the sub-agents work on these tasks, you can continue preparing your presentation or discussing other topics. Within minutes, you'll have all the information you need consolidated and ready for your meeting.

Conclusion

Sub-agents transform how you work with AI assistants by enabling true parallel processing. By delegating tasks effectively, you can dramatically increase your productivity and handle complex workflows with ease. Start experimenting with sub-agents today to unlock the full potential of your OpenClaw setup.