open-zk-kb

Persistent memory for AI coding assistants

Your AI assistant forgets everything between sessions. open-zk-kb fixes that.

It gives your assistant a structured, searchable knowledge base it queries automatically — so your preferences, decisions, and context persist across every conversation.

How it works

  1. Install in one commandbunx open-zk-kb@latest
  2. Your assistant stores knowledge — decisions, preferences, patterns, procedures
  3. Next session, it searches first — relevant context surfaces automatically

Notes are Markdown files with YAML frontmatter. A SQLite index provides full-text search, with local vector embeddings for semantic matching. No API key needed. No cloud required.

Key features

  • Hybrid search — full-text + semantic embeddings, so results match meaning not just keywords
  • Atomic notes — one concept per note, typed (6 kinds) with lifecycle management (fleeting, permanent, archived)
  • Local-first — everything stays on your machine, no API keys required
  • Multi-client — Claude Code, Cursor, Windsurf, OpenCode, Zed
  • Human-readable — Markdown files you can browse, edit, and version control
  • Rebuild from files — database is an index; your .md files are the source of truth
  • MIT licensed — use it however you want

Quick start

bunx open-zk-kb@latest

The interactive installer adds the MCP server to your client and injects instructions that teach your assistant when and how to use the knowledge base.

Learn more

Posts

subscribe via RSS