Skip to main content
Krilo Labs
Krilo Ship

Deploy at the speed of thought.

An MCP-native deployment platform. One tool call: domain, server, SSL, code. Live in 90 seconds.

How it works
1

1. Connect MCP

Add Krilo Ship to your AI tool — Claude, Cursor, or any MCP client. One config line.

2

2. Describe the deploy

Tell your agent the domain, repo, and region. No dashboard. No clipboard.

3

3. Agent ships

Krilo registers the domain, provisions a VPS, terminates SSL, deploys code, and verifies health.

4

4. Live in 90 seconds

Get a URL back. Iterate. Re-deploy with one tool call. Rotate credentials automatically.

Install via MCP

Install via MCP

claude_desktop_config.json json
{
  "mcpServers": {
    "krilo-ship": {
      "command": "npx",
      "args": ["-y", "@krilolabs/ship-mcp"],
      "env": { "KRILO_API_KEY": "sk_..." }
    }
  }
}
claude code bash
claude mcp add krilo-ship \
  --command "npx" \
  --args "-y,@krilolabs/ship-mcp" \
  --env "KRILO_API_KEY=sk_..."
Frequently asked questions
Model Context Protocol — an open standard from Anthropic that lets AI agents call external tools through a typed interface. Krilo Ship exposes deployment as an MCP tool.
Any MCP-compatible client. We test against Claude Code, Claude Desktop, and Cursor. Generic MCP SDKs (TypeScript, Python) work as well.
On a VPS provisioned in your chosen region (Singapore, US, or EU). You own the server. Krilo manages keys, secrets, and rotation.
No. Krilo issues short-lived credentials per deployment and rotates them automatically. There are no static SSH keys to leak.
Yes. Krilo can register a new domain via integrated registrars or use one you already own — point your nameservers at Krilo and the rest is automatic.
Those are designed for humans clicking buttons. Krilo is designed for AI agents calling tools. The whole API is structured, idempotent, and machine-first.
The previous revision keeps serving traffic. Krilo returns a structured error your agent can interpret and retry, escalate, or roll back without human intervention.
Yes. We collect deployment metadata (domain, region, status). We do not access your application code or runtime data beyond what is required to deploy and verify it.