Agent-Controlled Forgetting

A context management mechanism for AI agents that allows the model to actively select and compress historical tool-call results while archiving the full original content in a recoverable store. This reduces token consumption while preserving the ability to restore information when needed.