Optimization layer for Claude Code that reduces token spend 34–51% on average without changing your workflow or output quality.
Fermat's Last Token by Quotient Labs is a local wrapper and lightweight proxy that sits between the Claude Code CLI and the Anthropic API, compressing context, optimizing tool calls, and pruning older turns while preserving quality. It uses semantic context pruning, optimized agent tools that do more per call, and quality-preserving compression of tool outputs. Across 15 paired sessions on real SWE-Atlas refactoring tasks, Fermat was cheaper in every run and improved average quality score from 0.83 to 0.92. Supports metered billing and flat-rate unlimited plans with org-level access control. A 2-week free trial is available. Fermat's Last Token is a product of Quotient Labs.
For people
For agents
Nothing listed yet.