Concepts

The ideas that make aiflow effective with Claude Code:

• Memory: graph + RAG — why aiflow pairs a structural code knowledge graph (graphify) with semantic RAG (cocoindex) plus durable Beads memory, and how it routes questions.
• Token optimization — caveman terse output, rtk CLI-output filtering, graph/RAG retrieval, and cheap/local model routing — measured with aiflow cost.
• Team collaboration — a shared Dolt issue graph over your git remote, atomic claiming, and pull-before-push for many members.
• Features & advantages — the capability map and the case for the framework.

---

Table of contents

• Memory — graph + RAG
• Token optimization
• Team collaboration