Authless

Create and run a remote MCP server that lets you generate podcast-like URLs via Cloudflare AI and manage them from a lightweight MCP client. This server exposes two practical tools you can use to generate podcast URLs and retrieve recent entries, all without client-side authentication in the obvious demo setup. Use it to explore automated URL slug generation, lightweight content creation workflows, and simple podcast URL cataloging with cloud-backed storage.

How to use

You connect to your deployed MCP server from an MCP client or the Cloudflare AI Playground to run its tools. The server provides two core capabilities: the first tool generates a podcast-related URL slug based on a topic and returns the full Cloudflare Worker URL with that slug; the second tool lists recent podcast URLs and topics stored in the cloud database.

Typical usage patterns include: create a new podcast topic, invoke the generate_podcast tool to obtain a ready-to-use URL, and then list_recent_podcasts to review recent entries. These steps let you quickly compose, organize, and reuse podcast-like URLs for testing or demonstration purposes.

How to install

Prerequisites: you need Node.js and npm installed on your development machine.

1. Install the MCP project scaffolding using the provided bootstrap command.

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

2. Open a terminal in the project directory and install dependencies if needed. (The bootstrap step typically handles dependencies for you.)

3. Start or deploy the MCP server to a live URL as described in your chosen deployment flow.

Configuration and usage notes

The server is designed to be deployed at a Cloudflare Worker URL and accessed through standard MCP client workflows. A typical remote URL you may see after deployment looks like: http://remote-mcp-server-authless.your-account.workers.dev/sse. You can also access a development/no-auth variant via a local or remote HTTP endpoint if you provide the exact URL in your client and config.

To connect local clients or tools that support MCP, you’ll reference the server’s endpoint in your client configuration. For example, a configuration entry might point to the server’s sse endpoint used for streaming endpoints.

Examples and notes

Connect to Cloudflare AI Playground to experiment with MCP tools directly against your deployed server by entering your MCP server URL in the Playground interface.

If you want to run the MCP server locally with Claude Desktop or another client, you can configure an adapter that runs a local command and points to your remote sse endpoint. The exact command and arguments depend on your environment and client.

Available tools

generate_podcast

Takes a topic and makes an LLM call to Cloudflare Workers AI to generate a complex path slug related to the input topic, returning the combined Cloudflare Worker URL with that slug.

list_recent_podcasts

Takes a number and queries the Cloudflare D1 SQL database for previous podcast URLs and topics generated, returning a list of recent entries.