Skip to main content

Connect your agent

SpaceDock has one interface: tool calls. An agent deploys a directory, reads the errors, looks at the rendered page, queries the database and promotes a fork without a shell, a pipeline or a person in between. There is no build step and no image to push. Files are injected into a shared base image that is already running.

Setting that up is three steps, once. The last one is a paste.

  1. Get an account. Sign up with an email and a password. Where self-serve signup is closed (it is gated on the isolation milestone, and the sign-in screen says so in the server's own words), an operator mints you a key instead, and pasting that key is how you sign in.
  2. Press “Connect your agent”. It is on your dashboard while you have no apps, and on Account ever after. It mints a key and hands back the finished command with the key already in it. Nothing to splice together.
  3. Run it, then restart your agent's session. MCP servers are read at startup, so a session that is already open will not see it.

The line you get looks like this, with a real key where the placeholder is:

claude mcp add spacedock -e PLATFORM_API_URL=https://api.spacesagents.com -e PLATFORM_API_KEY=sk_… -- npx -y spacedock-mcp

Anywhere that is not Claude Code, the same server as an entry in the agent's own config:

{
  "mcpServers": {
    "spacedock": {
      "command": "npx",
      "args": ["-y", "spacedock-mcp"],
      "env": {
        "PLATFORM_API_URL": "https://api.spacesagents.com",
        "PLATFORM_API_KEY": "sk_9f2c…"
      }
    }
  }
}

Nine tools appear in the agent's session, and that is the end of your involvement. Everything under Reference is written for the agent. None of it asks you to run a command.