openclaw - 💡(How to fix) Fix Feature Request: subclaw - Independent Child Agent for Per-Directory Task Processing [1 comments, 2 participants]

Feature Request: subclaw - Independent Child Agent for Per-Directory Task Processing

Summary

I propose adding a subclaw feature to OpenClaw that would allow users to launch independent child agents focused on processing tasks within specific directories. This would mirror how Claude Code opens independent TUI windows as separate processes when launched independently, enabling parallel processing of tasks across different projects.

Motivation

Currently, all OpenClaw operations occur within a single main agent, which can lead to:

• Task interference when working on multiple projects simultaneously
• Performance bottlenecks when handling complex operations across different codebases
• Information pollution between unrelated projects

The proposed subclaw feature would solve these issues by creating isolated execution environments per directory.

Proposed Implementation

When a user navigates to a specific directory and runs subclaw, the following should happen:

1. A new, independent child agent launches
2. This agent focuses exclusively on tasks related to the current directory
3. The agent maintains its own context and state separate from the main OpenClaw instance
4. Like Claude Code's independent TUI windows, this would operate as a separate process/thread

Additionally, the feature should support project-specific configurations similar to Claude Code's .claude folder and CLAUDE.md file approach:

• Each directory could contain its own configuration files
• Context and state would be isolated per project
• Resources and processing power could be allocated independently per agent

Benefits

• Parallel Processing: Handle multiple projects simultaneously without interference
• Resource Isolation: Prevent resource contention between different projects
• Context Management: Maintain separate context per project directory
• Scalability: Enable distributed workload handling across multiple agents
• Clean Separation: Avoid information pollution between unrelated projects

Usage Example

cd /path/to/project1
subclaw  # Launches child agent focused on project1

# In another terminal
cd /path/to/project2
subclaw  # Launches separate child agent focused on project2

# Both agents operate independently with their own contexts

This feature would greatly enhance OpenClaw's capability for handling multiple simultaneous projects efficiently, similar to how Claude Code's independent TUI windows provide separate processing contexts.