Skip to content
Anthropic Engineering · Frontier Labs

Effective harnesses for long-running agents

Agents still face challenges working across many context windows. We looked to human engineers for inspiration in creating a more effective harness for long-running agents.