context-compression

You can reduce token usage in long agent sessions by applying context compression to summarize history while preserving critical decisions, file references, and next steps. This prevents context window bloat during extended workflows.

Anchored iterative summarization is a context compression technique that uses explicit sections to structure summaries over multiple passes. This prevents information drift by maintaining fixed reference points throughout the summarization process.

To summarize long conversations for code exploration and debugging workflows, trigger a structured compression pass at the session boundary to produce an anchored summary that retains critical decisions and file references.

Yes, context compression supports probe-based evaluation alongside a three-phase workflow (research, planning, implementation) to calibrate summary quality and ensure preserved context fidelity.

Use opaque compression for maximum token savings when artifact fidelity is less critical. Choose regenerative full summary instead when you need readable, structured refreshes of the compressed context.