Why I built another funnel builder when ClickFunnels, Funnelish, and CartFlows already exist

The honest reason: the closed-source incumbents are slow, expensive, and lock you in. The AI-native, open-source reason: even if they weren’t, the cost structure of building landing pages has changed enough that the category needs a rebuild.

The landscape, in three categories

Before I talk about Autonnel, here’s where the market sits in 2026, in three rough buckets.

The classic closed SaaS - ClickFunnels. $297/month gets you a hosted page builder, a CRM, an affiliate engine, and an email tool, all owned by one vendor. The product works. It also hasn’t fundamentally changed in years, the export options remain limited, and your funnels live exclusively on their infrastructure. If their billing system has an outage, your funnels go down with it.

The Shopify-native option - Funnelish. A narrower, cheaper, Shopify-only product. If you’re already on Shopify and just want one-click upsells without paying ClickFunnels prices, Funnelish has been a reasonable answer. But it’s still closed-source, still hosted on someone else’s infra, and still bound to a single commerce platform.

The WordPress plugin - CartFlows. Lives inside WooCommerce. You pay the WordPress tax (hosting, security patches, plugin compatibility hell) and you get a perfectly capable funnel builder grafted onto a CMS that was never designed for it.

Each of these works. None of them is what I want to build on in 2026.

What’s missing

Three things, in order of how much they bothered me.

1. Open source. Every funnel I’ve ever built has been worth, at some point, more than the software I built it on. The asymmetry of running your business on closed software you can’t fork is one of those things that feels fine until the vendor changes pricing, gets acquired, or decides to deprecate the API you depend on. I want a funnel builder I can read end-to-end, modify when I need to, and host wherever I want.

2. AI-native, in the actual sense. “AI-native” is one of the most overloaded terms in software. What I mean by it: the AI isn’t a feature, it’s the default authoring surface. You don’t draw a page and then ask a chatbot to suggest copy - you describe the page you want, and the system assembles it from real components, with real props, that you can then refine. The page generation, the conversion analysis, the A/B variant suggestions - all of it sits inside the loop, not next to it.

3. Adapter pattern at every boundary. Funnel builders historically pick a side: Shopify-only, WooCommerce-only, their-own-checkout-only. I want one product that doesn’t care which commerce platform you use, which payment processor you accept, which email service you send through, or which ad platform you report conversions to. Pick the stack you want, swap parts when you change your mind.

What Autonnel actually does

Concretely:

• Page generation. You describe a funnel (“4-step funnel for a wireless-earbuds DTC brand, video hero, color picker, PayPal Express checkout, post-purchase upsell”) and the AI assembles Puck-based pages with the right components wired up.
• Conversion analysis on a schedule. A cron job runs every 30 minutes to 24 hours (configurable), runs your recent funnel data through your LLM, and pings email, Slack, or a webhook with whatever the model thinks is slipping.
• Adapters everywhere. Shopify, WooCommerce, and our own open-source Picocart for commerce. Stripe, PayPal for payments. Resend, SMTP, and AWS SES for email. Facebook, TikTok, Bing Ads, and Google Ads for conversion postback. Swap one without touching the others.
• Runs anywhere. Node or Cloudflare Workers. Same code, same Astro Integration. Self-host on your laptop or run it at the edge.
• Apache-2.0. Fork, modify, ship. Use it commercially without asking.

What Autonnel isn’t trying to be

A few things I’m explicitly not building, in case it helps clarify the shape:

• Not a CRM. Order data lives in your commerce platform (Shopify, Woo, Picocart). Autonnel reads it; it doesn’t replace it.
• Not an email marketing tool. Transactional emails (order receipt, shipped, refunded, recall) are first-class. Drip campaigns and broadcast newsletters are not.
• Not an affiliate platform. Different problem space; not in scope.
• Not a no-code site builder for general use. The components and templates are specifically tuned for ecommerce funnels - landing, checkout, upsell, thank-you. If you want a generic WYSIWYG site builder, there are better tools.

Why now

Two things changed at roughly the same time:

LLMs got good enough to generate real components, not lorem ipsum. When the model can output structured component props that compile, the authoring surface flips. The page builder is no longer the bottleneck; the prompt is.

Edge runtimes got cheap enough to host funnels for free. Cloudflare Workers’ free tier covers most small DTC funnels indefinitely. Self-hosting an entire funnel stack used to mean a $20/month VPS plus a Postgres bill plus the time to patch all of it. Today it’s a wrangler deploy and a $0 monthly cost until you have real traffic.

Those two changes together mean the math of “build vs. buy” for a small ecommerce founder has flipped. You don’t need ClickFunnels pricing to get a working funnel stack anymore.

The bet

The bet behind Autonnel is straightforward: the next generation of funnel builders will be open-source, AI-native, adapter-driven, and runtime-agnostic. Someone is going to build that. I’d rather build it in the open with the people who’d use it than wait for one of the incumbents to figure it out.

If you’ve gotten this far and disagree with any of it, open an issue or start a discussion. The point of building in the open is hearing where I’m wrong.

Since this was written, the licence side of that bet has its own page: open-source funnel builders compared by licence, and the head-to-head with the incumbent is at Autonnel vs ClickFunnels. Running it costs nothing to self-host.

• github.com/autonnel/autonnel · Issues and feedback welcome.