title: Connect an agent to Ployz Cloud description: Give Claude, Codex or another MCP app the Ployz tools with a link, no CLI or token to install.

An agent app that speaks MCP can use the Ployz tools straight from Ployz Cloud. You add one link, sign in once in your browser, and pick the organization the agent works in. Nothing to install on your computer, and no token to copy.

This works on ployz.dev. A Self-hosted Cloud doesn't serve it; use ployz mcp there. ployz mcp has its own credential. Run ployz mcp login once to give it an agent token, which can't answer an approval.

Connect

1. Add https://ployz.dev/mcp to your agent as a remote MCP server. In Claude Code:

   sh

   1	claude mcp add --transport http ployz https://ployz.dev/mcp

   In other apps, look for Add custom connector or Remote MCP server and paste the link.

2. The app opens Ployz in your browser. Sign in if you aren't already.

3. Pick the Organization the agent works in and click Allow. The one you're working in now is picked to start with.

The agent now acts as you in that organization. To use another organization, connect again and pick it.

What it can do

The same tools as the Ployz agent in the dashboard. It can create projects, change services, configs, volumes and domains, save and deploy, follow and redeploy deployments, and work on your servers and volumes. Tell it which project and environment to work in.

If a reply gets lost, the agent can resend a write without running it twice, as long as it sends the same request_key (or x-ployz-request-key header) both times. A call without one is a new run each time, and Ployz tells agents this when they connect.

Approve a deploy

Every deploy or redeploy the agent asks for waits for you, even one that destroys nothing. A save waits only when it would remove something. See Destructive actions ask first. The call fails with approval_required, anything it would remove, and a link to your organization.

1. Open the link. The approval waits at the top of the Ployz agent under Waiting on you.
2. Read what the deploy does, and click Approve or Deny.
3. Tell your agent to try again. It sends the call with the approval, and the deploy goes ahead.

A service stop or restart, and removing or forgetting a server, wait the same way. A server drain, removing a project or environment, and releasing a volume wait when they would take something off your servers. None of this waits in an Agent sandbox nobody has claimed yet.

Until you answer, trying again only asks again, and nothing changes. A denied approval stays denied. Only you approve, in the dashboard. The agent can't approve its own request, and ployz org approval approve refuses it too with forbidden, before it asks for the organization's slug, with or without a terminal and in --json.

Disconnect

Log out of Ployz Cloud in the browser you connected from, and every agent you connected there stops working. Leaving the organization does the same for that organization. The next time the agent needs Ployz, it asks you to connect again.