How Deep Agents Manages Context: Offloading and Summarization
LangChain's Deep Agents harness uses offloading and summarization to combat agent context rot
“The more history, noise, and tool output a model carries, the worse its reasoning and accuracy tend to get. We call this context rot.”
LangChain details how their Deep Agents harness tackles 'context rot' — degraded model reasoning as agents accumulate history — via two mechanisms: offloading (routing large tool results and documents to a virtual file system, showing only previews) and summarization (compressing context at an 85% token threshold). The harness automates both techniques without developer intervention, framing context engineering as the core responsibility of an agent harness. This matters as a practical reference point for teams building production agents where long-running task quality degrades without explicit context management.