Softr Is Now Telling You to Build With Claude Code and Run on Softr — the 'Bring Your Own Coding Agent' Lane Just Got Another Tenant

Softr published a guide this week that reads like a small surrender and a big bet at the same time. The surrender: the company is now telling builders to do their actual building in Claude Code, Anthropic's AI coding agent, rather than inside Softr's own drag-and-drop editor. The bet: once Claude Code has produced your app, you run it, host it, permission it and manage it on Softr. That is the third lane of no-code emerging in public, and it is worth paying attention to because it tells you where the moat is moving.

What did Softr actually say?

On September 16 Softr published a piece framing the choice head-on: build with Claude Code, run on Softr. The positioning is blunt. Softr describes itself as the thing that turns a build into a hosted, permissioned business app you can put in front of users the same day. Claude Code, in Softr's telling, writes real source code inside your project but leaves the hosting, permissions and governance to you. Put the two together and you get the new pitch: let the coding agent do the construction, let Softr do everything after.

This has been building for a couple of weeks. Softr shipped an MCP integration in early September, which means you can drive Softr from Claude, ChatGPT and other AI tools, and it followed that with a workflow MCP and a vibe-code block MCP. The guide is the capstone. It is Softr saying, in public, that the editor is no longer the only place you build.

Isn't this Softr admitting drag-and-drop lost?

Sort of, and that is not the insult it sounds like. Drag-and-drop was never going to be how a builder handles a complex app, and everyone has known it for a while. What Softr is admitting is narrower and smarter: the hard part of building an app was never the visual editor, it was everything around the editor. Auth, roles, permissions, data, hosting, the stuff that turns a pile of generated files into something you can actually hand to a customer.

So yes, Softr is ceding the construction step to a coding agent. But it is doing it deliberately, because it thinks the construction step is becoming a commodity. The question is whether that bet is right, and I think it mostly is.

What is this third lane I keep hearing about?

Let me lay out the three lanes as I see them, because the distinction is starting to be useful.

• Lane one is drag-and-drop. Bubble, Glide, classic Softr. You build on a canvas, and the platform owns the whole stack. Still the default for most non-technical builders.
• Lane two is prompt-first. You describe an app in a chat and the tool generates it. Lovable, the early v0 wave, vibe coding. Fast to a demo, harder to production.
• Lane three is bring-your-own-agent. You build with Claude Code or another coding agent, then hand the output to a platform that hosts and governs it. That is Softr's new pitch.

Lane three is the interesting one, because it decouples who writes the code from who runs it. And once those two things are decoupled, the platform's moat stops being the editor and becomes the governance layer.

Why is the moat moving to the hosting layer?

Because editors are getting cheap and governance is not. A coding agent can generate a full app in an afternoon, but it does not give you auth, roles, permissions, audit trails, billing, or a safe place to run it. That is real, boring, expensive work, and it is exactly what platforms like Softr already have. When every builder can conjure code, the scarce asset is the part that makes the code safe to operate.

I think this is the correct read of where no-code is heading. The editor was the product for a decade. Now the editor is a feature, and the moat is everything that sits between generated code and a working, governed production app. Softr has just told the market it agrees, and it wants to be the place that catches whatever Claude Code throws at it.

There is an economics angle too, and it points the same way. Coding agents are priced by token or by work done, which means the cost of generation is falling fast. Governance is priced by seat and by month, and it is sticky. A company will churn its editor, but it will not churn the thing that holds its permissions and data. That is the kind of revenue a platform wants to anchor on.

Who else is in this lane?

Softr is not alone, which is why I am treating this as a pattern rather than a one-off. ToolJet has been pushing an MCP-native position for a while, letting external tools talk to its app platform directly. Microsoft has a canvas agent in Power Apps, which is the same idea from the enterprise side: describe it, let the agent build it, and the platform owns the lifecycle. Salesforce is doing a version of it too with its named agents and its long-horizon runtime.

The common thread is that the generation step is being commoditised, and everyone with an existing hosting and governance layer is racing to become the place where generated apps land. The editor wars are over, or close to it. The governance wars are just starting.

What should builders actually do differently?

First, stop treating your drag-and-drop platform as your identity. It is a tool, and the tools are converging. Second, start thinking about where your app lives and who controls it, because that is the decision that will be hard to reverse in two years. Third, and I mean this, try the lane before you argue about it. Spin up a small internal app with Claude Code and deploy it on Softr. See where it breaks.

My guess is the code generation will be the easy part and the governance will be where you spend your time, which is the whole argument in one experiment. If I am wrong, you lose an afternoon. If I am right, you just learned where your platform budget should actually go next year.

The takeaway

Softr telling you to build with Claude Code is not Softr giving up. It is Softr placing a bet that the editor is now a commodity and the governance layer is the moat, and I think that bet is right. The builders who win the next two years are the ones who stop obsessing over which editor to use and start treating the hosting, permissions and lifecycle layer as the thing they actually own. Everything else is about to get very cheap.