Add Deploy Target

Adding a Deployment Target

Overview

This MCP App uses a platform-agnostic app factory pattern. All business logic lives in api/app.ts, which exports a fetch handler. Each deployment target gets a thin entrypoint file (api/main.<platform>.ts) that starts the server using that platform's API.

When to Use

• User asks to "add Workers support", "deploy to Deno", "add Lambda entrypoint"
• User wants to deploy the app to a new platform
• User asks to add a wrangler.toml, Dockerfile, or platform config

Architecture

api/
├── app.ts              # Platform-agnostic core (DO NOT put platform code here)
├── main.bun.ts         # Bun entrypoint (default, used for local dev)
├── main.workers.ts     # Cloudflare Workers entrypoint (if added)
├── main.deno.ts        # Deno entrypoint (if added)
├── main.node.ts        # Node.js entrypoint (if added)
└── main.lambda.ts      # AWS Lambda entrypoint (if added)

Key rule: api/app.ts must never import platform-specific APIs. All platform specifics go in the entrypoint.

Step 1: Create the Entrypoint

Create api/main.<platform>.ts. The entrypoint imports app from ./app.ts and starts the server.

Cloudflare Workers

// api/main.workers.ts
import { app } from "./app.ts";

export default { fetch: app.fetch };

Deno

// api/main.deno.ts
import { app } from "./app.ts";

const PORT = Number(Deno.env.get("PORT")) || 3001;
Deno.serve({ port: PORT }, app.fetch);

console.log(`\nMCP App: http://localhost:${PORT}/api/mcp\n`);

Node.js

// api/main.node.ts
import { serve } from "@hono/node-server";
import { app } from "./app.ts";

const PORT = Number(process.env.PORT) || 3001;
serve({ fetch: app.fetch, port: PORT });

console.log(`\nMCP App: http://localhost:${PORT}/api/mcp\n`);

Requires dependency: bun add @hono/node-server

AWS Lambda

// api/main.lambda.ts
import { handle } from "hono/aws-lambda";
import { app } from "./app.ts";

export const handler = handle(app);

Requires dependency: bun add hono (for hono/aws-lambda adapter)

Step 2: Add Platform Config (if needed)

Cloudflare Workers — wrangler.toml

name = "mcp-app"
main = "api/main.workers.ts"
compatibility_date = "2025-04-01"
compatibility_flags = ["nodejs_compat"]

[build]
command = "bun run build:web"

[observability]
enabled = true

Add wrangler as a dev dependency:

bun add -d wrangler

AWS Lambda

Requires a bundling step (e.g., esbuild or bun build --target=node) and an infrastructure config (SAM template, CDK, or Terraform). Ask the user which IaC tool they prefer.

Deno

No config file needed. Optionally add a deno.json for import maps if the project uses npm packages:

{
  "imports": {
    "@decocms/runtime": "npm:@decocms/runtime"
  }
}

Step 3: Add Package Scripts

Add dev and deploy scripts for the new target in package.json:

Cloudflare Workers

{
  "dev:workers": "bun run build:web && wrangler dev",
  "deploy": "bun run build:web && wrangler deploy"
}

Deno

{
  "dev:deno": "deno run --allow-net --allow-env --allow-read api/main.deno.ts"
}

Node.js

{
  "dev:node": "node --experimental-strip-types api/main.node.ts"
}

Step 4: Verify

1. Run the new dev script and confirm the server starts
2. Test the MCP endpoint: curl http://localhost:3001/api/mcp
3. Run bun run check to ensure TypeScript is happy

Common Mistakes

Reference: Existing Entrypoint

The Bun entrypoint (api/main.bun.ts) is the reference implementation:

import { app } from "./app.ts";

const PORT = Number(process.env.PORT) || 3001;

Bun.serve({
  idleTimeout: 0,
  hostname: "0.0.0.0",
  port: PORT,
  fetch: app.fetch,
});

const slug = process.env.WORKTREE_SLUG;
const baseUrl = slug ? `http://${slug}.localhost` : `http://localhost:${PORT}`;

console.log("");
console.log(`MCP App: ${baseUrl}/api/mcp`);
console.log("");