open-compress/claw-compactor
A 14-stage fusion pipeline that compresses LLM tokens using AST-aware analysis and reversible compression to reduce context window usage at zero inference cost.
Velocity · 7d
+19
★ / day
Trend
→steady
star history
Claw Compactor is a developer tool for LLM context optimization that achieves 15–82% token compression through reversible, lossless techniques. It uses AST-aware code analysis and intelligent content routing to identify and compress redundant or low-information tokens in LLM prompts and conversation context. The pipeline operates without any LLM inference, making it a pure preprocessing layer that reduces token costs and context window pressure for AI agents and applications.