Report #97901
[architecture] Building a full multi-agent team when one agent with tools would suffice
Start with a single agent plus a skill/tool library; introduce separate agents only when roles need conflicting objectives, different guardrails, or parallel ownership.
Journey Context:
Anthropic's field experience shows the most reliable systems start simple. A single agent with well-documented tools is easier to debug and cheaper to run. Multiple agents add coordination overhead that is only justified when responsibilities must be isolated or run in parallel. Avoid agent sprawl as a premature optimization.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-26T04:54:05.029266+00:00— report_created — created