Welcome
repo-knowledge is a local-first knowledge system built on SQLite and FTS5. It catalogs every repository in your portfolio with structured metadata — thesis notes, architecture docs, audit evidence, inter-repo relationships — and exposes everything through a CLI and an MCP server for AI-integrated workflows.
What it does
Section titled “What it does”- Catalogs repos from GitHub orgs and local directories with tech fingerprints, topics, and docs
- Stores knowledge as typed notes: thesis, architecture, conventions, warnings, next steps
- Maps relationships between repos: depends_on, related_to, supersedes, shares_domain_with
- Runs audits with 80 controls across 19 domains, tracking findings, metrics, and posture over time
- Searches everything with FTS5 full-text search across READMEs, changelogs, notes, and descriptions
- Serves AI agents via 19 MCP tools that let Claude query, annotate, and audit repos conversationally
Who it’s for
Section titled “Who it’s for”Teams and solo developers managing 10+ repositories who want a single place to understand what they have, how it’s connected, and where the risks are.
Architecture
Section titled “Architecture”All data lives in a single SQLite database running in WAL mode. The system has two layers:
- Knowledge catalog — repos, tech fingerprints, docs, notes, facts, relationships, FTS search
- Audit evidence — structured audit runs, 80 canonical controls across 19 domains, normalized findings, metrics
The CLI (rk) and MCP server share the same database module and query layer.