Mercury adds parallel agents for coding

Parallel sub-agents are a meaningful step beyond chatbot-style assistants, turning Mercury into a workflow coordinator for complex development tasks.

• –Separate context windows let research, implementation, and testing agents stay focused
• –Automatic delegation and CPU/RAM-aware concurrency reduce manual task orchestration
• –Reader-writer file locks help prevent agents from overwriting one another’s work
• –Commands such as `/agents` and `/code agent` give developers visibility and direct control
• –Parallelism only pays off for independent tasks; coordination overhead, token costs, and shared-state conflicts remain the trade-offs