mex

name	mex
description	Drive mex (`mex-agent`), persistent project memory and code graphs for AI coding agents. One command scaffolds a living wiki, builds a deterministic code graph, and installs a project anchor file (CLAUDE.md, root AGENTS.md, .cursorrules, .windsurfrules, copilot-instructions.md, or .opencode/opencode.json) that your agent auto-loads as a standing rule document. Use when the user wants to `mex setup` a new project, build a symbol-grounded wiki, keep knowledge connected to implementation, route relevant context to agents, or run drift detection (`mex check`, `mex sync`). Triggers on: "mex setup", "project memory", "code graphs", "codebase documentation", "drift detection", "agent memory", "structured scaffolds", "architectural context", "living wiki", "project anchor file".
allowed-tools	Bash Read Write Edit Glob Grep
compatibility	Node.js >= 22.5. `npm install -g mex-agent` (MIT); binary is `mex`. Works standalone. An MCP package exists in the upstream monorepo (`packages/mex-mcp`) but is not published to npm as of this writing — build-from-source only, no `mex mcp` CLI subcommand ships in the released package. Supports TypeScript, JavaScript, Python, and Rust code analysis.
metadata	{"tags":"mex, mex-agent, project-memory, code-graphs, drift-detection, documentation, agent-memory, living-wiki, tree-sitter, sqlite, cli, project-anchor","platforms":"Claude, ChatGPT, Gemini, Codex, Cursor, Windsurf, jeo, gjc, jeopi, OpenCode","version":"1.1","source":"https://github.com/mex-memory/mex"}

Mex — persistent project memory for AI coding agents

Mex maps your code, turns what agents learn into structured Markdown, and keeps that knowledge connected to the implementation it describes. Every coding session starts with relevant architectural context instead of another full-repository scan.

When to use this skill

• Setting up a new project scaffold for an AI-assisted codebase (mex setup)
• Building a symbol-grounded living wiki tied to exact source locations
• Getting a project anchor file (AGENTS.md, CLAUDE.md, .cursorrules, ...) auto-generated so an agent picks up mex context as a standing rule, with no manual wiring
• Keeping architectural notes, decisions, and conventions synchronized with code
• Running drift detection (mex check) to catch stale documentation
• Syncing the wiki back to code after meaningful changes (mex sync)
• Retrieving compact, task-scoped context instead of a full-repo grep (mex graph scope "<task>")
• Migrating an existing pre-0.7 project into graph grounding (mex graph ground)

When not to use this skill

• The project has no meaningful history or architecture worth preserving as living documentation
• Documentation is generated purely from source comments (docstrings only, no structured memory needed)
• The user only wants an MCP server right now — mex's MCP package (packages/mex-mcp) is not published; do not promise MCP integration (see Compatibility above and Step 5 below)
• Real-time code inspection is more urgent than architectural context — mex complements, not replaces, LSP/IDE tools

Instructions

Step 1: One-shot auto-install (recommended)

bash .agent-skills/mex/scripts/install.sh /path/to/project

This is the fully automated path: it registers the skill, installs mex-agent globally if mex is not already on PATH, verifies that mex actually resolves to mex-agent (some machines already have a mex command — see the compatibility note below — the script refuses to proceed rather than silently running the wrong binary), runs mex setup non-interactively (skipped if .mex/ already exists — pass --force to re-run), builds the code graph with mex graph, runs for a health report, and prints which project anchor file mex installed. It is idempotent and safe to re-run. See for flags (, , , , , ).