Small changes shouldn't wait on engineering.

Marketing, design, and product describe the change in plain English. ronda writes the code and opens a pull request. Engineering reviews and merges.

No credit card required
Use it inChatGPT(opens in a new tab)Claude, coming soon

What is ronda

Plain-language code changes, shipped as real pull requests

ronda is a tool that lets non-engineers safely change a real web app. Connect a GitHub repository and describe the change you want in plain English. ronda applies it in an isolated live sandbox running your actual dev server, then opens a standard GitHub pull request your engineering team reviews and merges.

Works on your real repo

AI app builders generate throwaway projects from scratch. ronda edits the codebase you already ship, so there is no new framework to learn and no migration.

Every change is a reviewable PR

Nothing reaches production unreviewed. Each edit lands as a branch and pull request, so engineers stay in control and merge through their normal workflow.

Driven from the AI you already use

Run ronda from ChatGPT, Claude, or Cursor over MCP. The assistant does the editing, so there is no Anthropic key to add. It works as one tool in your agent's chain.

From request to merged PR without the ticket queue

Marketing edits without a ticket

A product marketer rewrites hero and pricing copy against the live preview and opens a pull request. An engineer merges it the same day, with no Jira ticket and no Figma round-trip.

Designers ship the polish

A designer adjusts spacing, button styles, and design tokens on the real app and shares the preview link for sign-off. The change lands as a reviewable diff instead of a redline handed back to engineering.

Engineers keep their focus time

Small label, copy, and layout tweaks that used to interrupt engineers now arrive as PRs from the rest of the team. Engineers review and merge them, and get back the focus time those interruptions used to cost.

From idea to merged PR in three steps.

A/B test, branch, and ship without engineering on call.

Beyond the describe-it-see-it loop, ronda gives non-engineers two tools usually reserved for product teams.

A/B experiments

Run real A/B tests in one prompt

Ask for a split-test in plain English. ronda writes both variants, wires the feature flag in your repo, and creates the experiment in PostHog. The pull request lands ready to merge. You click launch when you want traffic to start.

Connects to your PostHog. Results flow back into the dashboard you already use.

Side-by-side variants

Try two or three takes on the same change

Ask for variants instead of one answer. ronda generates each version on its own branch and lets you switch between them in the preview. Pick the winner and the others are cleaned up automatically.

Taste-driven iteration for marketing, design, and brand decisions where analytics would take weeks.

Drive ronda from the AI you already use

ronda plugs into Claude, Cursor, ChatGPT, and anything else that speaks MCP. Your AI can work alongside its other tools and push edits straight to a PR.

MCP server URLhttps://mcp.useronda.com
Claude
…

Or paste useronda.com/skill.md into agent tools that consume a skill file.

Start free, upgrade when you outgrow it

Free

Best for trying ronda end-to-end on a single change.

$0/month
Start free
  • 3 projects
  • 1 sandbox at a time
  • 1 collaborator
  • Use from ChatGPT or Claude, or the web editor with your own Anthropic key
  • Side-by-side variant generation
  • A/B experiments via PostHog

Starter

Daily-driver capacity. Your work stays ready to resume for two weeks between sessions.

$9/month
Get started
  • 5 projects
  • 5 sandboxes per project
  • Unlimited collaborators
  • Use from ChatGPT or Claude, or the web editor with your own Anthropic key
  • Side-by-side variant generation
  • A/B experiments via PostHog

Need higher limits? Contact us.

Using the web editor? You need an Anthropic key. How to get one. It's a 2-minute signup.

Before you choose

Before you choose ronda: the details teams ask about

Straight answers to what teams ask before they sign up: how you point at a change, what the AI costs, which apps work, and where your code and keys go.

Point at it, or describe it

ronda is not prompt-only. In the live preview of your real app, click Select element and point at the exact button, heading or card you want to change. ronda passes that element to the edit, so the change lands in the right component. You can also attach up to 3 images (a screenshot, a mockup, a Figma export) to show what you want.

ronda is not a drag-and-drop page builder, on purpose. Every change is written as normal code in your existing components and styles, so the pull request is something an engineer can read and merge. Nothing is layered on top of your app.

What the AI costs

Plans are Free ($0) and Starter ($9/month, unlimited collaborators). ronda never marks up AI usage. In the web editor you connect your own model key, from Anthropic or any OpenAI-compatible endpoint, and pay your provider directly. From ChatGPT or Claude no key is needed at all, because the assistant does the editing.

What ronda works on, and what it doesn't

  • Works with web apps that start with a Node dev server: Next.js, Vite, Astro and similar, including monorepos (pnpm, npm or yarn workspaces) and apps that live in a subfolder.
  • Not a fit yet: setups where the frontend needs a separate backend service started with docker-compose (each preview runs one app in one sandbox), and non-web repos such as mobile apps, Electron apps or smart contracts.
  • Check before you sign up: the free Repo analyzer scores how well your public repo fits, with no account needed.

Where your code and keys go

  • Each session runs your app in its own isolated sandbox VM. Nothing runs on your infrastructure and nothing is deployed.
  • ronda commits to a new branch and opens a pull request when you ask for one. It never merges, never pushes to your default branch and never deploys, so your branch protection and reviewers stay in charge.
  • Your model key is encrypted at rest with AES-256-GCM, and only its last four characters are ever shown back to you. Other stored credentials are encrypted the same way.
  • In the web editor, your instructions and the relevant code go to the model provider you chose, using your key. From ChatGPT or Claude they go to the assistant you use, under its provider's terms. ronda does not sell your data. Details in the privacy policy.

ronda or Builder.io Fusion?

Pick Builder.io if your designers want a full visual canvas, Figma-to-code and design-system indexing. Pick ronda if you want the shortest path from "please change this" to a reviewed pull request on the app you already ship, with point-and-click targeting, at $9/month for the whole team. Read the full ronda vs Builder.io comparison.

Frequently asked questions

What is ronda?

ronda lets you change a real web app by describing what you want in plain English. You connect a GitHub repository, ronda runs your actual app in a live sandbox, applies the change, and opens a pull request your team reviews and merges.

Can I use ronda inside ChatGPT?

Yes. ronda is an app in ChatGPT's app directory. Add it, sign in, connect your GitHub repository, and ask ChatGPT to change your site in plain English. You see a live preview, and when you ask for a pull request your team reviews and merges it. ronda never merges or deploys.

Do I need an Anthropic key to use ronda from ChatGPT?

No. When you use ronda from ChatGPT, ChatGPT does the editing, so there is no Anthropic key to add. Only the web editor uses a model key (Anthropic or any OpenAI-compatible endpoint), which you connect in your own workspace.

Do I need to know how to code to use ronda?

No. You describe changes in everyday language, like "make the headline bolder" or "change the pricing copy". ronda writes the code and your engineers review the pull request, so non-technical teammates can ship safely.

Can I point at the element I want changed?

Yes. In the live preview of your app, click Select element and point at the exact button, heading or card you want to change. ronda passes that element to the edit, so the change lands in the right component. You can also attach up to 3 images, such as a screenshot or a mockup, to show what you want.

How is ronda different from Lovable, v0, or Bolt?

Those tools generate brand-new apps from scratch. ronda works on your existing production codebase. It edits your real repository and ships changes as standard GitHub pull requests, so there is no rewrite and no new framework to adopt.

Does ronda work with my existing codebase?

Yes. Connect a GitHub repo and ronda boots your real dev server in an isolated sandbox, then edits your actual source files. It supports React and Next.js, Vue and Nuxt, Astro, and more.

What kinds of apps does ronda work with?

ronda works with web apps that start with a Node dev server, such as Next.js, Vite and Astro, including monorepos and apps in a subfolder. It is not a fit yet for frontends that need a separate backend started with docker-compose, or for non-web repos such as mobile apps, Electron apps and smart contracts. The free Repo analyzer at /tools/repo-analyzer scores your public repo before you sign up.

Is ronda safe to use on a production app?

ronda never pushes to production. Every change happens in a disposable sandbox and lands as a pull request on a branch. Your team reviews and merges it through the same workflow you already use.

Can ronda set up A/B tests?

Yes. Connect your PostHog workspace, then describe the test in plain English. ronda writes both variants, wires the feature flag, opens a pull request, and creates the experiment in PostHog. You decide when to launch traffic. For quick taste tests without analytics, ask for variants and pick the winner in the preview.

How much does ronda cost?

ronda has a free plan at $0 with three projects, one sandbox at a time, and one collaborator. The Starter plan is $9/month with five projects, five sandboxes per project, and unlimited collaborators. Using ronda from ChatGPT or Claude needs no key, because the assistant does the editing. In the web editor you connect your own model key (Anthropic or any OpenAI-compatible endpoint) and pay your provider directly for AI usage.

Can I use a model other than Anthropic in the web editor?

Yes. A workspace admin can connect Anthropic or any OpenAI-compatible endpoint with their own key, and edits are sent to that provider. You pay the provider directly and ronda never marks up AI usage. From ChatGPT or Claude no key is needed.

Which AI assistants can drive ronda?

ronda is an app in ChatGPT: add it from ChatGPT's app directory and sign in. It also exposes its actions over the Model Context Protocol (MCP), so Claude, Cursor, or any MCP-capable agent can drive it. You can also paste useronda.com/skill.md into tools that consume skill files.

How long does it take to get started?

It takes under five minutes. Sign in with GitHub, connect a repository, and describe your first change. No credit card required.

Contact Sales

Ready to accelerate your team?

Talk to an expert about how ronda can transform your development workflow.