⏱️ Time MCP Server on Cloudflare Workers (Remote, No Auth)

This project is a remote MCP server running on Cloudflare Workers that gives LLMs rich time awareness:

• Current time (UTC + any timezone)
• Relative time (“3 hours ago”, “in 2 days”)
• Days in month
• Timestamps
• Timezone conversion
• Week / ISO week of the year

It is based on Cloudflare’s remote-mcp-authless template, but exposes a full time toolkit over MCP.

---

Tools

The server exposes these MCP tools:

• current_time
  Get the current time in UTC and in a target timezone, in a configurable format.

• relative_time
  Get the relative time from now for a given timestamp (e.g. "2025-03-23 12:00:00" → "in 2 days").

• days_in_month
  Get the number of days in a given month (or the current month if no date is provided).

• get_timestamp
  Get the Unix timestamp (milliseconds) for a given time, or for now.

• convert_time
  Convert a given time between two IANA timezones (e.g. Europe/Amsterdam → Asia/Tokyo).

• get_week_year
  Get the week number and ISO week number for a given date, or for today.

All tools are implemented in src/index.ts using this.server.tool(...) on a McpServer instance.

---

Get Started

You can either:

1. Deploy from this repo (once it’s on GitHub)

Use the Deploy to Workers button (update the URL to match your GitHub repo):

This will deploy your MCP server to a URL like:

https://mcp-timeserver.<your-account>.workers.dev/sse

2. Create locally from the Cloudflare template

If you want to reproduce this setup from scratch:

npm create cloudflare@latest -- mcp-timeserver --template=cloudflare/ai/demos/remote-mcp-authless

Then replace src/index.ts with the time MCP implementation from this repo and install dayjs:

npm install dayjs

---

Running Locally

From the project directory:

npm install
npm run start

This starts a local dev server, usually at:

http://localhost:8787

Exposed endpoints:

• SSE MCP transport: http://localhost:8787/sse
• HTTP (streamable) MCP transport: http://localhost:8787/mcp
• Health check: http://localhost:8787/health

---

Deploying to Cloudflare

To deploy the Worker:

npx wrangler login
npm run deploy
# or
npx wrangler deploy

After deploy you’ll get a URL like:

https://mcp-timeserver.<your-account>.workers.dev

With MCP endpoints:

• SSE: https://mcp-timeserver.<your-account>.workers.dev/sse
• HTTP: https://mcp-timeserver.<your-account>.workers.dev/mcp

---

Using the MCP Server

1. Cloudflare AI Playground

You can use this Worker as a remote MCP server from the Cloudflare AI Playground:

1. Go to https://playground.ai.cloudflare.com/

2. Add a new MCP server and use your deployed URL, for example:

   https://mcp-timeserver.<your-account>.workers.dev/sse
   

3. The tools current_time, relative_time, etc. will be available for the model to call.

---

2. MCP Inspector (local testing)

To inspect tools and make test calls:

npx @modelcontextprotocol/inspector@latest

In the Inspector UI:

• Use http://localhost:8787/sse (dev) or your deployed /sse URL.

• Connect and then List Tools to see all time tools.

• Call tools like current_time with JSON arguments, for example:

    {
      "format": "YYYY-MM-DD HH:mm:ss",
      "timezone": "Europe/Amsterdam"
    }
  

---

3. Connect Claude Desktop via mcp-remote

Most desktop clients still speak stdio MCP, so you can use mcp-remote as a proxy to your Cloudflare Worker.

In your Claude Desktop config (e.g. claude_desktop_config.json):

{
  "mcpServers": {
    "time-mcp-timeserver": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp-timeserver.<your-account>.workers.dev/sse"
      ]
    }
  }
}

Restart Claude. You should now see:

• current_time
• relative_time
• days_in_month
• get_timestamp
• convert_time
• get_week_year

available as tools.

---

Development Notes

• Runtime: Cloudflare Workers + Durable Objects
• MCP: @modelcontextprotocol/sdk via Cloudflare agents/mcp
• Time library: dayjs with utc, timezone, relativeTime, weekOfYear, isoWeek plugins.
• Entry point: src/index.ts
• Config: wrangler.toml

---

License

MIT – feel free to fork, extend, and adapt.