Documentation

Installation

Pick the Milkey integration path that matches your workflow and get connected in minutes.

Milkey supports two production-ready entry paths. If you are using a coding assistant like Cursor or Claude Code, start with the hosted MCP setup. If you are integrating Milkey into an application, use the official TypeScript SDK.

Choose Your Path

Hosted MCP: best for Cursor, Claude Code, Codex, and other MCP-compatible agents.
Official SDK: best for TypeScript products that want Milkey-managed skills inside existing OpenAI, Anthropic, Gemini, or AI SDK stacks.

Get API Key

Every Milkey integration starts with an API key.

Open the API Keys section in your dashboard.
Click Create Key.
Copy the key immediately and store it in your environment or secret manager.

Keep Secrets Server-Side

Never commit Milkey keys to source control or expose them in browser code. The key should live in environment variables, secret managers, or server-only configuration.

Connect Milkey

1{2  "mcpServers": {3    "milkey": {4      "url": "https://mcp.vexelityai.com/mcp",5      "headers": {6        "Authorization": "Bearer mk_sk_your_api_key_id_your_secret"7      }8    }9  }10}

1import { milkey } from "@milkeyskills/sdk"2 3const milkeyClient = milkey.createClient({4  baseUrl: process.env.MILKEY_BASE_URL!,5  apiKey: process.env.MILKEY_API_KEY!,6})7 8const tools = milkey.openai.chat.tools({9  client: milkeyClient,10})

If you are using the SDK, continue with the Milkey Skills SDK guide. If you need agent-specific MCP steps, start with the Cursor guide or use the generic MCP overview from the MCP section.

Verify Setup

MCP: confirm the agent reports the Milkey server as connected.
SDK: run a minimal tool registration flow and verify the provider client receives Milkey tools.