A few years ago, we started microfeed, an open-source project for building self-hosted CMS on Cloudflare. It’s kind of like WordPress, but without the hassle of managing your own servers.
The original motivation was simple: give podcasters a way to host their shows on Cloudflare for FREE, with up to 10 GB of free media storage on Cloudflare R2 and support for their own custom domain.
But microfeed quickly became useful for much more than podcasting. You can use a microfeed instance to host a blog, photo album, PDFs, or pretty much any collection of content you want to publish on the web. We at Listen Notes run several microfeed instances for US $0 on Cloudflare:
Listen Notes Changelog - changelog.listennotes.com
Brand assets - brand-assets.listennotes.com
Our mascot llamacorn - llamacorn.listennotes.com
Fast-forward to 2026. AI agents have become genuinely useful, so naturally, we’re evolving microfeed to work well with them too.
First, try it
Open your coding agent - Codex, Claude Code, or another agent with command-line access, and give it this prompt:
Deploy microfeed to Cloudflare.
Start by running
npx @microfeed/cli manage, then follow its instructions until deployment is verified.
The agent will walk through the setup and deploy your own microfeed instance. Here’s a video demo:
You can use it to host a blog, podcast, photo album, or other content. Cloudflare R2 includes 10 GB of free storage, which is enough for roughly 300 hours of podcast audio or 10,000 images.
If you’d rather use the old deployment workflow, that’s still available too: you can deploy microfeed to Cloudflare through GitHub Actions.
Once your instance is up and running, there are a few things in the new microfeed worth exploring.
Content automation
microfeed now comes with a command-line tool, @microfeed/cli, that both you and your AI agents can use.
The CLI can customize a deployed microfeed instance, publish content, and perform administrative tasks directly from the command line. This makes it possible to automate quite a bit of the work involved in running a content site.
Here’s a real example.
The Listen Notes changelog website is a microfeed instance. When we want to publish a new changelog entry, we can simply ask an AI agent - in this case, Codex - to create an unlisted post about a new product feature on changelog.listennotes.com:
The agent writes the post, publishes it to the changelog site as unlisted, and gives the URL to a human editor for final review and approval.
microfeed itself becomes the content infrastructure the agent can operate.
Every microfeed instance also comes with an API and webhooks, so it’s easy to connect it to Zapier or another automation system.
For example, you could set up a workflow like:
New post published → send a tweet → send an email newsletter
The API lets you create, update, and retrieve content from your microfeed instance. Webhooks work in the other direction, pushing events when content is created, updated, deleted, and so on.
The new /admin dashboard also makes it much easier to experiment with the API and webhooks.
And the /admin dashboard supports WebMCP, which means an AI agent can interact with the dashboard on your behalf to help with content management.
Website customization
The new microfeed also comes with a theme system for customizing how your site looks.
There are seven built-in themes to choose from. You can activate one as-is, or create your own version and edit its HTML, CSS, and JavaScript directly in the browser.
For more substantial changes, you can export a theme to a Git repository, ask a coding agent to work on it, and then import the updated theme back into your microfeed instance. Here’s a git repo of the theme used by www.microfeed.org - www-microfeed-org-theme.
You can also create and edit site-level files such as robots.txt, llms.txt, and ads.txt directly from the /admin dashboard.
In particular, files such as llms.txt can be useful for providing instructions and context to AI agents interacting with your site.
Making self-hosting less tedious
Self-hosting usually comes with a price: operations.
You have to think about backups, custom domains, migrations, storage, configuration, and all the other mundane tasks that come with running your own site.
The @microfeed/cli tool makes a lot of these tasks easier—especially now that you can simply ask a coding agent to perform them for you.
Coding agents are quite good at figuring out which commands and command-line arguments they need.
For example, you can ask an agent to download the entire D1 database and R2 bucket behind your microfeed instance for backup, or use that data to migrate your content to a new instance.
The goal isn’t to pretend that self-hosting requires zero operations. It’s to make those operations simple enough that you or your AI agent can handle them without turning site maintenance into a project of its own.
Give it a try
microfeed started as a way to make self-hosted podcasting cheap and simple - you buy your own domain name, but spend $0 in hosting fees.
It’s becoming something broader: a self-hosted publishing system that both humans and AI agents can operate.
You own the site. You own the content. You can use your own domain. And now, increasingly, you can delegate the tedious parts to an AI agent.
Give it a try: