Review

Floot and Cursor in 2026: Build in Cursor, Ship with Floot

Adrian Yumul
Adrian Yumul• Published Mar 18, 2026• Updated
Floot and Cursor in 2026: Build in Cursor, Ship with Floot

Floot and Cursor come up together a lot, usually framed as a choice. They aren't one. Cursor is where the code gets written; Floot is what the code runs on. You can point one at the other and keep working in the editor you already like.

This guide covers how that works, what it changes about the parts of a project that usually stall, and when Cursor on its own is still the right answer.

What Cursor Does

Cursor is an AI-powered code editor built for developers — completions as you type, refactors on a highlighted block, multi-file context across a codebase, an agent that can run longer tasks, and code review through Bugbot. You're still the one writing and reviewing the code; Cursor makes that faster.

It's also a full MCP client. Cursor's agent can connect out to external MCP servers and call real tools through them, with OAuth for servers that need it, in the editor and in the Cursor CLI both. That's the hook everything below hangs on.

What Floot Does

Floot is the app platform: a real Postgres database, authentication, serverless endpoints, file storage, custom domains, and publishing to the web, with mobile app conversion on the Power plan.

You drive it through a connector. Floot speaks the standard remote MCP protocol over Streamable HTTP with OAuth, so any client that supports remote MCP servers can use it — Cursor, VS Code, Zed, the Codex CLI, and the ChatGPT and Claude apps.

Floot also has its own in-app editor and its own AI agent for people who'd rather work that way. That path spends Floot credits. Driving Floot from Cursor doesn't.

Connecting Floot to Cursor

Add Floot as an MCP server in Cursor, authorize it, and Cursor's agent picks up Floot's tools: read and write project files, run SQL against the database, typecheck, and deploy.

From then on it's the ordinary Cursor loop. You describe a change or write it yourself, the agent applies it, and when it's ready you say ship it — no context switch, no second window, no separate deploy tool.

What This Actually Changes

The interesting part isn't the editor. It's the work that comes after the code, which is where a lot of Cursor projects quietly stop.

Left to yourself, that list is: pick a host, provision a database and keep it patched, wire up auth, manage secrets and environment variables, build a deploy pipeline, point a domain at it, run migrations, and fix production when it breaks. Then do the deploy part again on every subsequent change, forever.

With Floot underneath, that list is already there and already running. There's no separate "now deploy it" step — the edit and the publish are the same act, on the tenth change as much as the first. Your agent has the database on the other end of a tool call, so a schema change and the code that depends on it can land together.

Every AI editor is good at producing code now. Far fewer things take that code and make it a running app with a URL, a database and real users on it.

How the AI Gets Billed

Your Cursor billing doesn't change. Cursor's default is a subscription with usage-based overage, and it also supports bring-your-own-key — supply your own OpenAI, Anthropic, Google or Bedrock key, available even on its free tier, and pay that provider directly. Either way, that arrangement is between you and Cursor.

Floot doesn't add a second AI bill on top of it. Some app builders work by handing your prompt to their agent, which does the building on their infrastructure and spends their credits — so you pay your editor's AI bill and theirs for the same work. Floot's connector doesn't do that. It hands your agent primitives and your agent does the reasoning, so there's no second AI in the loop to pay for.

What Floot meters is the platform work: actions — the edits, reads and tests your agent runs against the project — and hosting. The action allowance is sized for a full day of heavy building and refills every morning whether you used all of it or none of it, and a failed action isn't counted, so telling your agent to try again after a bad result costs nothing.

Your Cursor Session Comes With It

Because your own agent is doing the building, everything else you've set up in Cursor applies. Other MCP servers you've connected — a Figma server, an issue tracker, internal tools — are available in the same breath as Floot's. Your rules files, your model choice and the conversation you've already had about this project all carry over.

A vendor's agent on the far end of a connector sees none of that. It gets the prompt that was forwarded to it and nothing else.

Where the Code Lives

Worth being straight about: a Floot project's files live in Floot, and Cursor's agent reads and writes them through the connector rather than from a folder on your disk. You get the editor and the agent you're used to, driving a codebase that's hosted rather than local.

If owning the repo is the point — you have an existing codebase, custom infrastructure, or a stack that has to be yours — that's a real reason to use Cursor by itself, and no amount of convenience downstream changes it.

When Cursor Alone Is the Right Call

Use Cursor on its own if you're extending an existing codebase, if your architecture needs decisions Floot doesn't make for you, or if managing the infrastructure is part of the job rather than a tax on it. It's a strong tool for that, and pairing it with Floot isn't an upgrade to it — it's a different shape of project.

When the Pair Makes Sense

Reach for both when you want a real, hosted app at the end of it and the infrastructure isn't the interesting part: internal tools, client portals, dashboards, booking systems, marketplaces, anything with logins and a database behind it. You keep writing and reviewing code in Cursor. Floot takes it from there.

Getting Started

  1. Create a Floot account and a project.
  2. Add Floot as an MCP server in Cursor and authorize it.
  3. Enable Floot's tools in Cursor and describe what you want built.
  4. Ship it from the same conversation.

The same connector works from the ChatGPT and Claude apps if you'd rather build by chatting some days and stay in the editor on others — it's the same project either way.

Final Thoughts

The question was never Floot or Cursor. Cursor is an editor, and one of the good ones. Floot is the database, the backend and the deploy that an editor doesn't give you.

Put them together and you write code the way you already do, and stop losing days to everything that comes after it.

Adrian Yumul

Adrian Yumul

Floot and Cursor in 2026: Build in Cursor, Ship with Floot - Floot