Find the ghost tokens. Fix them. Survive compaction. Avoid context quality decay.
Measure token savings per AI coding agent, optimize context, and share a live local knowledge graph across 16 CLI clients.
Cut the tokens Claude Code loads at session start. CLI that sets up a small CLAUDE.md with on-demand docs, measures startup tokens, and keeps them small.