← Back to Discover

Tree Ring Memory

Open-source Rust command-line tool that stores what AI agents should remember, such as decisions and lessons, in a local SQLite database with search, auditing and a terminal UI.

Command-line tools
Visit website ↗

What it does

Tree Ring Memory is a local-first memory layer for AI agents. Agents can keep useful decisions, warnings, preferences and lessons without storing whole transcripts. Fresh memory stays detailed, older memory is compressed into rings, important warnings stay visible, and durable truths become heartwood. It is a Rust command-line tool with SQLite and full-text search storage and no required cloud service. Commands cover remembering and recalling memories, import and export, audit checks, consolidation and maintenance, and a terminal UI for browsing the store. Privacy defaults block memories that look like secrets, and capturing whole transcripts is not part of the tool. The tool also connects to agent setups: it can configure project-local integrations, offers an optional instruction-only plugin for Codex and Claude Code, and includes adapters for the DOX and Revolve patterns it draws on. The project describes itself as being in protocol-preview status.

Capabilities

  • Local SQLite store with full-text recall and no required cloud service
  • Memory is aged into rings, with a way to keep durable truths
  • Audit, consolidation and maintenance commands
  • Import and export of memories
  • Terminal UI for searching and browsing the store
  • Privacy defaults that block secret-like memories

When to use it

  • You want AI agents to carry decisions and lessons across sessions without keeping full transcripts
  • You need agent memory kept on your own machine with no cloud dependency
  • You work across several agents on one host and want a shared, auditable memory