Headroom: How Context Compression Cuts Agent Token Costs by 60–95% Without Changing Answers

Headroom sits between agents and LLMs as a compression layer, shrinking tool outputs, logs, RAG chunks, and files before they hit the context window.