context isolation
Running a piece of work in its own separate context, such as a sub-agent, so only its result returns to the main conversation.
In more detail
Isolation keeps the main context short and free of raw tool output, but it costs extra tokens because each context carries its own instructions, and the sub-agent cannot see what the main agent knows unless its brief says so. Use it when work splits cleanly and the raw material is bulky.
Related terms
- sub-agent — An agent started by another agent to do one piece of work in its own fresh context, returning only its finding.
- context engineering — Choosing and arranging everything a model reads on each call, from instructions and tools to history, memory and retrieved data, within a token budget.
Taught in
- When Context Fails — AI Engineering: LLM Apps, RAG & Agents
- Planning, Memory And Many Agents — AI Engineering: LLM Apps, RAG & Agents
References
- Anthropic: How we built our multi-agent research system
- Drew Breunig: How to Fix Your Context (2025-06-26)