@vallor/openwork-agent
v1.2.0
Published
Initialize and manage AI agent configuration for any project
Maintainers
Readme
openwork-agent
Initialize and manage AI agent configuration for any project. One .agents/ directory, shared across Claude Code, Cursor, Gemini CLI, and Codex CLI.
Install
npm install -g openwork-agentRequires Node.js >= 18 and a global ~/.agents directory with your skills, commands, and hooks.
Usage
openwork-agent init [dir]
Set up .agents/ in a project directory. Walks you through selecting skills, commands, hooks, MCP servers, and AI tools — then symlinks everything from your global ~/.agents.
openwork-agent init # current directory
openwork-agent init ./myapp # specific directory
openwork-agent init --all # skip prompts, link everythingopenwork-agent add <url>
Install a skill from a GitHub repository into ~/.agents/skills (and optionally symlink it into the current project).
openwork-agent add user/repo
openwork-agent add https://github.com/user/repo
openwork-agent add user/repo --skill my-skill # skip pickeropenwork-agent mcp [name]
Add or manage MCP servers in .agents/mcp.json.
openwork-agent mcp # interactive picker
openwork-agent mcp slack # add Slack directlyHow it works
openwork-agent init creates a .agents/ directory in your project, then symlinks .claude, .cursor, .gemini, and .codex to it. Skills, commands, and hooks are symlinked from your global ~/.agents, so changes propagate everywhere.
your-project/
.agents/ # real directory
skills/ # symlinks → ~/.agents/skills/*
commands/ # symlinks → ~/.agents/commands/*
hooks/ # symlinks → ~/.agents/hooks/*
mcp.json # shared MCP server config
AGENTS.md # project-specific instructions
.claude → .agents
.cursor → .agents
.gemini → .agents
.codex → .agents
CLAUDE.md → .agents/AGENTS.md
.mcp.json → .agents/mcp.jsonLicense
MIT
