Long, tool-heavy conversations can repeatedly send large amounts of context back to a model. Moxby reduces eligible repeated weight while preserving the source text needed for recovery.
Tool Condenser
The Tool Condenser creates a compact representation of verbose tool output. The conversation keeps enough information to continue the work, while the full output remains available when exact details need to be recovered.
Context Rasterization
Context Rasterization can carry eligible, stable context in a denser visual form. Moxby keeps exact strings and source text available for retrieval rather than relying on a visual rendering for values that must be copied precisely.
Context restoration
When you return to a workflow, Moxby can restore relevant browser and conversation context instead of requiring you to explain the work again. This can extend the useful life of a conversation and reduce repeated token weight.
How savings are reported
The usage report shows measured token reductions and estimated value for eligible activity. Results depend on the model, conversation, cache behavior, and tool output. Plan allowances describe available capacity, not a guaranteed amount of savings.
Read the Context Rasterization technical paper for the design, measurement approach, and limitations.