Anthropic says one AI agent beats many subagents now that models can handle complexity alone

Why it matters

Anthropic is advising enterprises to consolidate fleets of specialized sub-agents into a single primary agent that loads context on demand, and has published a migration kit on GitHub to guide the process. The recommendation stems from its Applied AI team's work across enterprise deployments, where multi-agent designs produced higher latency, rising costs, and information loss at handoffs. Newer models such as Opus 4.5, Opus 5.5, and Fable 5.1 can now handle complex planning within a single loop, doing orchestration work themselves. Anthropic describes how organizational structure gets mirrored in agent architecture, leading to overlapping agents and contradictory system prompts. It recommends a clean system prompt, a small set of tools, skills loaded on demand, and a central team owning quality. Sub-agents remain sensible for parallel tasks or when context windows degrade, and Anthropic cautions against merging blindly after one test broke a working process. The findings align with outside research questioning whether agent swarms improve quality.