EN
English简体中文繁體中文日本語EspañolBahasa Indonesia
Get started

Claude Code plugin · open source

Ship, fix and roll back by asking your agent

The Zenavo skills teach Claude Code how to deploy your projects, find out why a deploy failed, go back to a working version and manage environment variables. It works through a key you scope, and asks before it changes anything.

InstallView on GitHub

This page is also available in 简体中文, 繁體中文, 日本語, Español and Bahasa Indonesia.

Claude CodeExample

Why did the last deploy of api fail?

zenavo-debug

  • get_project shop
  • list_deployments api
  • get_job
  • list_env api

The image built, but api exits on start: it reads DATABASE_URL, and the service has no variable by that name. Add it in the service's Variable tab so the value stays out of this chat, and I'll deploy again.

Install in three steps

Once per computer. The key stays on your machine, never in Claude Code's settings.

  1. Install zenavo-mcp

    The skills talk to Zenavo through zenavo-mcp, an open-source MCP server. Download the binary for your system and check it (the README shows how), or build it with Go 1.26+:

    go install github.com/Zenavo-App/zenavo-mcp/cmd/zenavo-mcp@latest

    Download from Releases →

  2. Sign in and scope the key

    zenavo-mcp login

    The console opens and asks you to approve. Check that the code matches your terminal, choose Deploy or Read only, and limit the key to the projects the agent works on. To only let it look, sign in with --scope read.

  3. Add the plugin to Claude Code

    claude plugin marketplace add Zenavo-App/zenavo-skillsclaude plugin install zenavo@zenavo

    The plugin brings the five skills below and starts zenavo-mcp itself.

Five skills

Each one is a short, readable playbook in the zenavo-skills repository. Claude picks the right one from what you ask.

  • zenavo-deployAsks you first

    Deploy a project or one service, follow the job, and check it came up healthy.

    Deploy the api service of shop

  • zenavo-debug

    Find out why a deploy failed or a service crashes, from status, job output and logs, and propose a fix.

    Why does the worker keep restarting?

  • zenavo-rollbackAsks you first

    Go back to an earlier working deployment of a service or a whole project.

    Roll web back to the previous version

  • zenavo-envAsks you first

    Add, change or remove environment variables, keeping the values out of the conversation where possible.

    Set LOG_LEVEL to debug on the worker

  • zenavo-setup

    Install and sign in to zenavo-mcp when the tools are missing or not signed in.

    The Zenavo tools are not working

Other agents

Any MCP client can use the same server, with the same key and the same limits. The skills are made for Claude Code; elsewhere the agent gets the tools.

Claude Code, server only

If you only want the tools, without the plugin:

claude mcp add zenavo -- zenavo-mcp

Claude Desktop, Cursor and others

Add the server to the client's MCP configuration (claude_desktop_config.json, .cursor/mcp.json, …):

{
  "mcpServers": {
    "zenavo": { "command": "zenavo-mcp" }
  }
}

What the agent can call

The tools zenavo-mcp gives the agent. What a key can do is enforced by the Zenavo API, not by the agent: a Read only key cannot deploy.

Read

  • whoamiThe account, teams and roles the key acts with, and its permission
  • list_projectsProjects and their services
  • get_projectServices, ports, domains, live state, the active job and deploy blockers
  • get_logsThe newest log lines of a service, optionally filtered
  • list_deploymentsRecent deployments of a service, marking the current one
  • get_jobA job's status, progress and log; can wait for it to finish

Deploy permission

  • list_envNames of a service's environment variables, never their values
  • deployDeploy a project or one service
  • rollbackRun an earlier successful deployment again
  • restart_serviceRestart a service without rebuilding
  • set_env / unset_envSet or remove one environment variable; the value is never shown back

Built to be handed a key

An agent with the Deploy permission can still deploy code that does anything your services can. So the key is narrow, and you stay in the loop.

  • It asks first

    Deploys, rollbacks, restarts and variable changes wait for your yes. Tools that change something are marked, so your agent asks you before running them.

  • Values stay hidden

    The agent sees variable names, never values. Logs reach it with the project's variable values already replaced, and marked as untrusted data.

  • A key you scope

    Deploy or Read only, limited to some projects, expiring after 90 days. It is saved on your computer, not in the agent's settings.

  • On the record

    Every change appears in your audit log under the key's name. Delete the key under Account settings → API Keys to cut it off at once.

There are no tools to delete projects, open a shell, read files or database rows, download backups, change servers, or touch billing and team members. Those stay in the console.