AI Briefing
KO

Kauri, a Decision Record Tool

·2026.04.25 04:22

Key point

Kauri has released a local-first tool that records, searches, and validates design decisions made by agents and humans on a Git basis.

Details

Kauri is a local-first decision record store used together by LLM agents and humans, storing architecture choices, rules, and constraints.

  • Records are committed together with the repo and version-controlled with git, and are injected into the agent context at the start of a session.
  • State is managed as draft / active / superseded / deprecated.
  • It supports file linking, staleness detection, full-text search, and a constrained tag taxonomy.

The usage is also concrete.

  • kauri init creates a repository
  • kauri record leaves a decision along with its rationale, tags, and file links
  • kauri project generates the projection to be given to the agent
  • kauri check inspects stale records

There are two scopes.

  • Project scope: stored in .kauri/store.db and shared with the team.
  • User scope: stored in ~/.kauri/store.db and reused as personal, cross-project standards.

Agent integration is also included.

  • It provides a plugin and MCP tool for Claude Code.
  • It can also be used with MCP clients such as Cursor, Windsurf, and Claude Desktop.
  • setup-git registers a custom merge driver to handle simultaneous edits across branches.

The CLI provides record, update, query, show, history, validate, pin, project, taxonomy, serve, and more, and all commands support --json and --help.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.