@an-sdk/cli

Deploy AI agents to AN from your terminal.

Quick Start

# 1. Login with your API key (get one at an.dev)
npx @an-sdk/cli login

# 2. Create your agent
npm init -y && npm install @an-sdk/agent zod

Create src/agent.ts:

import { agent, tool } from "@an-sdk/agent"
import { z } from "zod"

export default agent({
  model: "claude-sonnet-4-6",
  systemPrompt: "You are a helpful assistant.",
  tools: {
    add: tool({
      description: "Add two numbers",
      inputSchema: z.object({ a: z.number(), b: z.number() }),
      execute: async ({ a, b }) => ({
        content: [{ type: "text", text: `${a + b}` }],
      }),
    }),
  },
})

# 3. Deploy
npx @an-sdk/cli deploy

That's it. Your agent is live.

Commands

an login

Authenticate with the AN platform.

npx @an-sdk/cli login
# Enter your API key: an_sk_...
# Authenticated as John (team: my-team)

Your key is saved to ~/.an/credentials.

an deploy

Bundle and deploy your agent.

npx @an-sdk/cli deploy
# Bundling src/agent.ts...
# Bundled (12.3kb)
# Deploying my-agent...
# https://api.an.dev/v1/chat/my-agent

The CLI:

1. Finds your entry point (src/agent.ts, src/index.ts, agent.ts, or index.ts)
2. Bundles your code + dependencies with esbuild
3. Deploys to a secure cloud sandbox
4. Returns your agent's URL

How It Works

Your code  -->  an deploy  -->  Cloud Sandbox  -->  Clients
                                (E2B + Claude)

• Your agent runs in an isolated cloud sandbox with Node.js, git, and system tools
• The Claude Agent SDK executes your agent with the model and tools you defined
• Clients connect via SSE streaming at the returned URL

Project Linking

After first deploy, the CLI saves .an/project.json in your project directory. Subsequent deploys to the same project update the existing agent.

Environment Variables

AN_API_URL — Override the API endpoint (default: https://an.dev/api/v1)

License

MIT