MEX 🧩 Skill Open source
Team memory for engineers and their agents, kept in the repo and shared through Git
- GitHub stars
- 1.7k
- Stars this week
- +17
- Forks
- 139
- Licence
- MIT
- Last push
- 2026-10-02
- Maintainer
- mex-memory
npx [email protected] setupThird-party agent skills run with your permissions. Read the source before installing, and prefer pinned versions.
Works with
About MEX
What it does
MEX is shared project memory that lives inside your repository and travels through Git. It keeps a team's architecture notes, decisions, requirements and handoffs as readable Markdown next to the code, linked to the actual source through a code graph. Engineers and their coding agents read that context, propose additions or corrections for review, and pass work between sessions and teammates with structured handoffs called relays. There is no hosted service: each person keeps their own local indexes and a local browser Hub, and Git is the only sharing layer. It also works solo, where the next reader is you in a later session.
What is inside
- A Wiki of architecture, decisions, conventions and patterns, grounded in a code graph
- An Inbox where an agent drafts a knowledge addition or correction for a person to approve
- Relays that carry a handoff: what changed, which tests ran, what remains, and where to look next
- A local Project Hub on 127.0.0.1 for exploring knowledge, reviewing proposals and managing members
- A CLI (
mex) for setup, graph and wiki building, queries, and the inbox and relay workflows - Agent integrations that install instruction anchors and skills for your coding agent
Works with
Claude Code and Codex get skill commands. Cursor, Windsurf, GitHub Copilot and OpenCode get instruction anchors. It needs Node.js 22.5 or newer and a Git repository.
How to install
Run from the repository root:
npx [email protected] setup
Maintenance and safety
MEX is MIT-licensed and actively developed. Canonical files are committed and shared through your normal Git flow; the setup wizard only commits on your explicit request and never pushes or pulls. Pseudonymous CLI and Hub telemetry is on by default and can be turned off with mex telemetry disable, MEX_TELEMETRY=0 or DO_NOT_TRACK=1; names, paths and content are excluded. The Hub protects changes with a session and CSRF token.
Who should use it
It suits teams, and solo developers across sessions, who want durable project context and clean handoffs kept in the repo rather than in a separate tool. If you do not use Git or Node, it will not fit.
Pros
- Durable project context and handoffs, no hosted service
- Agents propose knowledge changes for human review
- Works solo and across teammates via Git
Cons
- Pseudonymous telemetry on by default (can be disabled)
- Needs Node.js 22.5+ and a Git repository
Similar agent skills
All development workflow skills →Superpowers 🧩 SkillFree
A skills library that makes coding agents plan, test-first and debug systematically
Matt Pocock's Skills 🧩 SkillOpen source
Small, composable engineering skills: grilling, TDD, specs, tickets and architecture review
Agent Skills (Addy Osmani) 🧩 SkillOpen source
25 lifecycle skills and commands that make agents spec, test, review and ship with discipline
Ponytail 🧩 SkillOpen source
Makes your coding agent write the minimum code a task needs, without cutting safety
Understand Anything 🧩 SkillOpen source
Turns a codebase into an interactive knowledge graph with tours, search and Q&A
Archify 🧩 SkillOpen source
Validated, interactive architecture, sequence and data-flow diagrams compiled from typed JSON