Summary
Add a toggle (per-project or global) that prevents Agent Zero from autonomously spawning new subagents at runtime, so a manually designed multi-agent system stays exactly as built.
Current Behavior
Agent Zero can spawn subagents autonomously during task execution based on its own judgment. There is currently no way to prevent this at runtime.
What Already Exists (and Why It Doesn’t Cover This)
I’m aware of the following existing controls:
• agents.json per project — enables/disables predefined subagent profiles
• Plugin toggle system — scoped ON/OFF for plugins via .toggle-1 / .toggle-0
• Agent profiles with settings.json — per-agent overrides for model, tools, context
None of these prevent the agent from spawning new subagents autonomously at runtime. They control which profiles exist, not whether autonomous spawning is allowed.
Requested Behavior
A toggle (UI and/or agents.json flag) with two states:
• Auto (default): Current behavior — agent spawns subagents as needed
• Locked: Agent cannot spawn new subagents autonomously. Only manually defined subagents are active. The system architecture stays frozen as designed.
Use Case
When building stable, intentional multi-agent pipelines — e.g. for production deployments, client-facing systems, or reproducible architectures — autonomous spawning at runtime introduces unintended structural changes. A lock mode would allow users to design a system once and guarantee it runs exactly as intended.
Notes
• Default behavior should remain unchanged (no breaking change)
• Ideally configurable both globally and per-project
• Could also be exposed as a system prompt instruction as a lightweight alternative
Summary
Add a toggle (per-project or global) that prevents Agent Zero from autonomously spawning new subagents at runtime, so a manually designed multi-agent system stays exactly as built.
Current Behavior
Agent Zero can spawn subagents autonomously during task execution based on its own judgment. There is currently no way to prevent this at runtime.
What Already Exists (and Why It Doesn’t Cover This)
I’m aware of the following existing controls:
None of these prevent the agent from spawning new subagents autonomously at runtime. They control which profiles exist, not whether autonomous spawning is allowed.
Requested Behavior
A toggle (UI and/or agents.json flag) with two states:
Use Case
When building stable, intentional multi-agent pipelines — e.g. for production deployments, client-facing systems, or reproducible architectures — autonomous spawning at runtime introduces unintended structural changes. A lock mode would allow users to design a system once and guarantee it runs exactly as intended.
Notes