commander

To dispatch parallel agents for independent tasks, you invoke a multi-agent orchestration tool that assigns one agent per domain, ensuring each operates with a fresh, isolated context to prevent cross-contamination while executing concurrently.

Multi-agent orchestration coordinates isolated agents working in parallel to reduce investigation time. You need it when facing 3 or more independent failures or research angles that can be executed concurrently without shared state.

No, isolated agents for concurrent tasks require non-inheriting session history. Each agent must operate with a fresh, isolated context and a clearly defined objective to prevent cross-contamination of context and ensure deterministic outcomes.

The best way to coordinate independent research angles is to dispatch one agent per domain with per-agent context construction. This approach maintains isolated context boundaries and uses coordinated result collection to aggregate findings.

You should not use parallel agents for multi-agent dispatch when your tasks require shared state, have dependent execution paths, or involve fewer than 3 independent failures or research angles that cannot run concurrently.