Squeez is an end-to-end token optimizer designed to significantly reduce token consumption for popular AI CLIs, including Claude Code, GitHub Copilot CLI, OpenCode, Gemini CLI, and OpenAI Codex CLI. It achieves this by compressing bash output by up to 95%, intelligently collapsing redundant calls, and automatically injecting a terse prompt persona. Operating with zero new runtime dependencies, Squeez enhances efficiency and lowers costs by ensuring that only essential context is passed to the language model, featuring advanced techniques like signature-mode for code reads, cross-call deduplication, and a self-teaching protocol.
Key Features
0174 GitHub stars
02Self-teaching protocol payload for LLMs to understand its markers and commands
03Integrated MCP server providing 13 read-only tools for session memory access
04Persistent session memory with structured summaries of prior work and next steps
05Cross-call context deduplication using exact-hash and fuzzy Jaccard matching
06Bash output compression up to 95% reduction via smart filtering and truncation
Use Cases
01Significantly reduce token costs for AI CLI usage with Claude Code, Copilot CLI, and other supported platforms.
02Enhance LLM context window efficiency by compressing redundant bash outputs and prior session data.
03Provide AI agents with structured access to session memory and historical context via an MCP server.
