# ronda > Edit your web app in plain English. ronda turns natural-language requests into precise code changes, runs them in a live sandbox running your real dev server, and ships them as GitHub pull requests your team reviews and merges. ronda lets non-engineers safely change a real web application. You connect a GitHub repository, describe the change you want in plain English, preview it live, and open a standard pull request. There is no rewrite and no new framework to adopt. AI app builders generate throwaway projects from scratch; ronda edits the codebase you already ship. ronda is bring-your-own-key (you use your own Anthropic API key and pay Anthropic directly) and can be driven from Claude, Cursor, ChatGPT, or any MCP-capable agent. Pricing: the Free plan ($0) includes 1 project, 1 sandbox at a time, and 1 collaborator. The Starter plan ($9/month) includes 5 projects, 5 sandboxes per project, and unlimited collaborators. Both plans are bring-your-own Anthropic key. No credit card is required to start. ## Product - [Homepage](https://useronda.com/): What ronda is, how it works, pricing, and FAQ - [Pricing](https://useronda.com/pricing): Free ($0) and Starter ($9/month) plan details, both bring-your-own Anthropic key - [FAQ](https://useronda.com/#faq): Common questions about ronda - [FAQ for LLMs](https://useronda.com/faq-for-llms.md): Flat, extractable question-and-answer reference about ronda, optimized for answer engines ## Compare ronda to alternatives - [Comparisons hub](https://useronda.com/compare): Side-by-side overviews of how ronda differs from Lovable, Bolt, v0, and a GitHub seat - [ronda vs Lovable](https://useronda.com/compare/ronda-vs-lovable): Lovable generates a new app from a prompt. ronda edits the app you already ship and opens a pull request for engineering review. - [ronda vs Bolt](https://useronda.com/compare/ronda-vs-bolt): Bolt builds and runs a new app in the browser. ronda changes the app already in your GitHub repo. - [ronda vs v0](https://useronda.com/compare/ronda-vs-v0): v0 generates UI from a prompt. ronda modifies the page that already exists and ships the diff as a reviewed pull request. - [ronda vs giving your PM a GitHub seat](https://useronda.com/compare/ronda-vs-github-seat): A GitHub seat plus an AI editor is friction and blast radius. ronda is one element, plain English, and a reviewed pull request. - [ronda vs Google Optimize](https://useronda.com/compare/ronda-vs-google-optimize): Google Optimize shut down in 2023. ronda sets up an A/B test on your real app in plain English and ships it as a pull request your team reviews, with no developer needed to build the test. - [ronda vs Base44](https://useronda.com/compare/ronda-vs-base44): Base44 generates a new app from a prompt. ronda edits the app you already ship in your GitHub repo and opens a pull request your engineers review. - [ronda vs Replit](https://useronda.com/compare/ronda-vs-replit): Replit builds, runs, and hosts an app in the cloud. ronda edits the app already in your GitHub repo and opens a pull request your engineers review. - [ronda vs Framer](https://useronda.com/compare/ronda-vs-framer): Framer is a visual website builder where the site lives in Framer. ronda edits a real coded app in your GitHub repo and ships changes as reviewed pull requests. - [ronda vs VWO](https://useronda.com/compare/ronda-vs-vwo): VWO applies A/B test variants as a runtime visual overlay. ronda writes the variant into your real code and ships it as a pull request your team reviews, with no flicker on React or Next.js and results in your own PostHog. - [ronda vs Optimizely](https://useronda.com/compare/ronda-vs-optimizely): Optimizely is an enterprise experimentation platform. ronda lets a non-engineer set up an A/B test in plain English without an SDK or a contract, and ships the variant as a reviewed pull request with results in your own PostHog. - [ronda vs Statsig](https://useronda.com/compare/ronda-vs-statsig): Statsig is experimentation built for engineers who integrate the SDK. ronda lets a non-engineer set up an A/B test in plain English and ships the variant as a reviewed pull request, with results in your own PostHog. ## Migrate from another tool - [Switching from Lovable](https://useronda.com/from-lovable): What to expect when moving an existing Lovable export onto ronda, plus FAQ on export, rebuild, stack, team, and cost. ## For non-technical teams - [Cursor for non-developers](https://useronda.com/cursor-for-non-developers): Cursor assumes you read diffs and run a repo. ronda lets a non-developer change the app you already ship in plain English, and every change lands as a pull request your engineers review. - [A/B testing on your real app](https://useronda.com/ab-testing): Set up an A/B test on your real React or Next.js app in plain English. ronda writes the variant into your code and opens a pull request your team reviews. The split runs in your code with no overlay flicker, and it uses your own PostHog project for assignment and results. - [ronda for product managers](https://useronda.com/for-product-managers): Most AI tools for PMs build throwaway prototypes. ronda lets a product manager change the real product in plain English and ship it as a pull request engineers review and merge. - [ronda for designers](https://useronda.com/for-designers): Stop handing off small UI tweaks to engineering. ronda lets a designer change the real production app in plain English, fixing spacing, font weights, colors, and copy, and ship it as a pull request engineers review. - [Edit code with AI without being a developer](https://useronda.com/edit-code-with-ai): Most AI code editors assume you write code. ronda lets a non-developer edit a real app's code in plain English and ships every change as a pull request engineers review. ## Free tools - [Free tools hub](https://useronda.com/tools): Index of ronda's free tools for getting an existing app ready for ronda, no account required - [Repo analyzer](https://useronda.com/tools/repo-analyzer): Paste a public GitHub URL and ronda scores how well its sandbox-and-edit workflow will fit your stack, with explicit signals for framework, dev command, Docker, i18n, and build complexity. - [Lovable migration kit](https://useronda.com/tools/lovable-migration): Pick your framework and ronda generates a step-by-step kit (commands, files, checklist) for moving a Lovable export into a normal repo. ## For AI agents - [MCP overview](https://useronda.com/mcp): How to drive ronda from Claude, Cursor, or ChatGPT over MCP, the agent opens a sandbox on your repo, makes a change, and opens a pull request your team reviews - [Skill file](https://useronda.com/skill.md): Instructions for an AI agent driving ronda over MCP - [MCP agent guide](https://useronda.com/mcp.md): Full reference for integrating with the ronda MCP server - [MCP server](https://mcp.useronda.com): Model Context Protocol endpoint ## Get started - [Sign up](https://app.useronda.com/signup): Create a workspace for free, no credit card - [Sign in](https://app.useronda.com/login): Existing accounts ## Writing - [Your PM shouldn't have to wait three days to change a button](https://useronda.com/blog/pm-shouldnt-wait-3-days-to-change-a-button): Why non-technical teammates can make web edits safely when the workflow stops being the wall. - [5 ways to tell if your AI writing tool is actually working](https://useronda.com/blog/ai-writing-tool-worth-keeping): Five concrete signals to measure whether your AI writing tool is helping, from acceptance rate to domain proximity. - [5 tasks Codex /goal can run while you sleep](https://useronda.com/blog/codex-goal-tasks-overnight): Five async Codex tasks worth queuing tonight, from inbox triage to test coverage, with goal templates that actually finish. - [The AI paradox: More automation, more humans, more work](https://useronda.com/blog/ai-paradox-more-automation-more-humans-more-work): Why every AI automation creates new human oversight work, and what it means for product managers and designers in 2026. - [When AI should show its work: 5 findings from a 450-person user study](https://useronda.com/blog/when-ai-should-show-its-work): A platform governance study found that showing arguments beats giving recommendations by 28 points when the AI can be wrong.