I posted mex here a few weeks ago, it crossed 700+ stars and outside contributors started shipping PRs. Just released v0.3 with a terminal dashboard, heartbeat checks, event logs, and agent-memory mode.

Hello! I posted about mex here a few weeks back and the response was honestly insane, first of all thanks. For anyone who wants to get to the real stuff straight away, links in the replies. Since then mex crossed 700+ stars, PRs started coming in from contributors I had never met, and I just released mex v0.3. What is mex? mex is a structured markdown scaffold that lives in `.mex/` in your project root. Instead of one giant context file, the agent starts with a tiny bootstrap file that points to a routing table. The routing table maps task types to the right context files. Working on architecture? Load the architecture context. Writing new code? Load conventions. Debugging? Load debugging notes. Need a repeatable workflow? Load patterns. The key idea is simple: the agent should load only the context it needs, not the whole damn project. In v0.2, mex was mainly a drift-aware scaffold CLI. It helped keep project memory accurate. v0.3 turns it into a lightweight operational memory layer for agents. there are loads of new things in this update, let me list out a few * Terminal dashboard: running `mex` now opens an interactive TUI with scaffold health, drift score, heartbeat status, recent events, and quick actions. * Agent-memory mode: `mex setup --mode agent-memory` creates a scaffold for persistent agents, with daily memory, task logs, decisions, heartbeat checks, and stronger GROW guidance. * Heartbeat checks: `mex heartbeat` checks whether memory is still fresh, including stale files and cleanup signals. The part I’m most excited about is the agent-memory mode. This is for workflows where the “project” is not just a codebase anymore. It could be a persistent local agent, a homelab, an OpenClaw-style operational workspace, Kubernetes/Docker/Ansible/Terraform runbooks, or any long-running context where the agent needs to preserve state over time. A nice way to frame it: mex v0.2 helped agents avoid stale project context. mex v0.3 helps agents maintain working memory over time. Install/update: npm install -g mex-agent@latest or: npx mex-agent@latest setup For agent-memory mode: npx mex-agent@latest setup --mode agent-memory mex heartbeat I’m still trying to make mex much better, especially for persistent agents and long-running AI workflows. If anyone here likes the idea and wants to contribute, please do. I’m actively reviewing PRs and trying not to make people wait. Once again, thank you.