Problem / Motivation
The current subagent system is closer to background task delegation than a true multi-agent system.
A subagent receives a task, executes it independently, and returns a final result to the main agent. However, subagents do not have persistent identities, shared task state, agent-to-agent communication, or the ability to coordinate and delegate work among themselves.
This limits more complex workflows that require multiple specialized agents to collaborate, share intermediate results, and coordinate dependent or parallel tasks.
Proposed Solution
Consider evolving the current design toward an optional multi-agent system where multiple agents can:
- have distinct roles and responsibilities;
- communicate and collaborate with each other;
- share task status and intermediate results;
- coordinate dependent or parallel tasks;
- delegate bounded subtasks when necessary.
The existing subagent feature could remain as the lightweight option for simple background delegation, while multi-agent coordination could support more complex workflows.
Alternatives Considered
Keep the current one-shot subagent model and rely on the main agent to manually coordinate all delegated tasks. This remains useful for simple background work, but it does not provide direct collaboration or shared coordination between agents.
Related Component
Agent / Prompts
Additional Context
This should follow nanobot's "core stays small; extend at the edges" principle. Ideally, multi-agent coordination could be introduced as an optional tool, plugin, or external service without adding significant complexity to AgentLoop or AgentRunner.
Problem / Motivation
The current subagent system is closer to background task delegation than a true multi-agent system.
A subagent receives a task, executes it independently, and returns a final result to the main agent. However, subagents do not have persistent identities, shared task state, agent-to-agent communication, or the ability to coordinate and delegate work among themselves.
This limits more complex workflows that require multiple specialized agents to collaborate, share intermediate results, and coordinate dependent or parallel tasks.
Proposed Solution
Consider evolving the current design toward an optional multi-agent system where multiple agents can:
The existing subagent feature could remain as the lightweight option for simple background delegation, while multi-agent coordination could support more complex workflows.
Alternatives Considered
Keep the current one-shot subagent model and rely on the main agent to manually coordinate all delegated tasks. This remains useful for simple background work, but it does not provide direct collaboration or shared coordination between agents.
Related Component
Agent / Prompts
Additional Context
This should follow nanobot's "core stays small; extend at the edges" principle. Ideally, multi-agent coordination could be introduced as an optional tool, plugin, or external service without adding significant complexity to AgentLoop or AgentRunner.