All blogs
Open‑source memory and guardrails for the coding agent you already use
A free command-line tool that gives Claude Code, Codex and other coding agents durable knowledge, memory and guardrails as plain markdown.
- Client
- h1902y
- Year
- 2026
- Status
- Open source
- Timeline
- Public since June 2026
The problem
Coding agents forget what they learned and repeat mistakes. Teams have no simple way to grow an agent’s knowledge safely over time.
What we built
- A command-line tool that sets up memory, actions and guardrails for the agent you already run
- Notes stored as plain markdown, grown through a loop where a human approves what the agent learns
- Tracing with OpenTelemetry, host hooks that gate risky actions, and a web workbench
- No runtime dependencies, MIT licence, published on npm
Stack
- Node.js
- TypeScript
- OpenTelemetry
- Markdown
Where it landed
- Built in public since June 2026, with over a thousand commits
- Works with Claude Code, Codex, Gemini CLI, OpenCode and pi
Want something like this?
Tell us the idea in a few lines. We will say what week one looks like.
