Replacing vector-search RAG with Markdown docs + knowledge graphs on Claude CLI
Shivam__kumar · reddit · 2026-11-07
The author building an AI system for a complex ERP app replaced vector-search RAG with a documentation-first architecture: structured Markdown files for every module, business rule and DB table relationship, a knowledge graph Claude generates from those docs, a read-only MCP server for live database queries, all wired through Claude CLI with custom rules. Results beat the old RAG setup, with an estimated 95% hallucination reduction (unbenchmarked), and no per-question handlers needed since Claude picks its own tools. Multi-user isolation is being enforced at the app/MCP-server level, not via prompts. The author stresses it's an experiment, not a claim that vector RAG is useless.
More from coding & agent
- Tip: Run /prototype on Small UI Changes to Get Several Options to Pick From — mattpocockuk · 2026-11-07
- Databricks VP of Engineering and Neon Co-founder on Building Databases for AI Agents — matei_zaharia · 2026-11-07
- Multi-model coding apps are being acquired by labs — the survivors are open source — petergyang · 2026-11-07
- One folder that Claude reads first: the underrated Claude setup for solo operators — PrajwalTomar_ · 2026-11-07
- Linear's AI code review tool Reviews now free on all plans, no Linear tracking required — julianlehr · 2026-11-07
- Stately ships a SKILL that has LLMs author agent state machines through 5 validation gates — DavidKPiano · 2026-11-07