Replit

The one that brings the machine with it

Every other tool in this section edits code on your computer. Cursor and Windsurf are editors, Codex and Claude Code are agents in your terminal. All four assume you have a repository, a runtime and somewhere to deploy.

Replit does not. The environment, the database and the hosting arrive with the agent, which makes it a different proposition rather than a different interface.

Three products, and the name has moved

Replit's documentation now describes three: Chat for research and planning, Build for creating apps and sites, and Design for interfaces.

If you are looking for "Replit Agent", the agent is what you talk to inside this structure rather than a separate product. Documentation still covers working with Agent directly, including Agent Skills for specialised knowledge, but the entry points are the three above.

Worth stating because the roadmap name and the product name have diverged, the same way they had for Codex and Windsurf. Older tutorials describe a layout that no longer matches the interface.

Prompt to published, without leaving

The documented path runs start from a prompt, build a small app, test it in Preview, then choose release settings and publish.

Nothing in that sequence involves your machine. The database is added from inside the project so data survives a refresh, restart and publish. Deployments are a few clicks. Development URLs let you share work in progress before any of it is released.

For a prototype that is the whole appeal. The gap between an idea and a URL you can send someone is usually infrastructure, and here that gap is closed by default.

Projects, and what persists

The Workspace gathers projects, chats, routines and data, and the Project Editor is where you talk to Agent. Chats can become Projects, so exploratory work has a path into something buildable rather than being thrown away.

Routines are the scheduled-work piece, which is the same idea Claude Code calls by that name. On a hosted platform it matters more, because there is no local machine that has to stay awake.

What you are trading

Three things, and they decide whether this is the right tool rather than a pleasant one.

You do not own the runtime. Hosting, scaling and the environment are the platform's. Convenient until you need something it does not offer, at which point you are migrating rather than configuring.

Publishing costs scale with the app. Deployment pricing is usage-based, so a prototype that succeeds has a bill that grows with it. Read that page before you put anything real on it.

Your workflow lives there. The project is not a git repository on your disk that happens to be edited by an agent. Moving to another tool later is an export, not a cd.

None of those is a flaw. They are the cost of not having to set up a machine, and for the audience that wants the app rather than the codebase, it is the right trade.

Who it is for

Reach for Replit when getting something running matters more than owning the pipeline: a prototype to test an idea, an internal tool nobody will maintain, a demo with a deadline, or learning where setup is the obstacle.

Reach for the local tools when the code has to live in your repository, pass your CI, and be maintained by people who need it on their machines. That is most production software, and it is why these tools coexist rather than compete.

Further reading

• Replit, Documentation: Chat, Build and Design, the publish flow, and deployment pricing.
• Cursor and Windsurf: the local agentic editors.
• Claude Code: the agent that runs where your code already is.
• Self-hosted models: the same own-it-or-rent-it question, one layer down.