evex documentation: the community registry for eve agents
Install community eve agents into your project with one command and own every file they write. What evex is and how the shadcn-compatible registry works.
What evex is
evex is the community registry for eve agents. Each agent is a source-owned package: its files, metadata, dependencies, and author identity live in a public GitHub repository, and every agent enters or changes the catalog through a reviewed pull request.
The registry speaks the shadcn registry format, so the same CLI that installs UI components installs agents. There is no upload form and no opaque bundle. What you see in the repository is exactly what an install writes into your project.
The evex database stores runtime state only, such as install counts and favorites. The canonical definition of every agent is the code in the repository, which means you can audit an agent before you trust it.
What installing gets you
Installing an agent copies its full source into your project. The files land under the agent/ directory layout that eve expects, alongside evals, a README, and an environment template when the agent needs credentials.
After the install finishes, evex is out of the loop. Your project has no runtime dependency on the registry: the agent runs from the files in your repository, you can edit any of them, and nothing phones home.
- Full agent source under agent/ in your project, readable and editable
- Evals, a README, and a .env.example template when the agent reads environment variables
- npm dependencies declared by the agent, prompted during install
- No runtime dependency on evex after the files are written
How evex relates to eve
eve is Vercel's framework for building AI agents. It defines the agent/ directory convention that holds an agent's configuration, instructions, skills, tools, and subagents.
evex is the distribution layer on top of that convention. It does not run agents and it is not a hosted platform: it packages community agents so any eve project can install them with one command and own the result.
Where to go next
The Installation page covers prerequisites, the install command, and what to do after the files are written. The Registry page documents the HTTP endpoints behind the catalog, including the machine-readable resources for tools and LLMs. The MCP page shows how to browse and install agents from editors like Cursor, VS Code, and Claude Code through the shadcn MCP server. The Publishing page walks through adding your own agent to the catalog by pull request.