cursor live apps: Vibe Coding Leaves Localhost

The cursor live apps release is the clearest sign yet of where vibe coding is heading. On November 5, Cursor published a changelog entry called Build Live Apps: its cloud agents can now build a web app from a prompt, provision a Postgres database, set up sign-in, host it, and hand back a shareable link, all inside the same editor that wrote the code. The tool that writes your code now wants to run it too. Source

Cursor's own framing is blunt: agents have lowered the cost of writing code, but most apps people build never make it off localhost. This release is an attempt to close that gap, and it lands directly in the territory that internal-tools platforms, Lovable, and Replit have been competing for. Here is what shipped, what the guardrails look like, and what it changes for founders building with agents. Source

What Cursor shipped

According to the changelog, you start a cloud agent and tell it explicitly that you want to build an app. The agent scaffolds the web application, spins up a Postgres database if the app needs persistent storage, configures authentication behind Cursor SSO, and provisions hosting. If the app needs to reach your other systems, like a data warehouse, Cursor asks for the secrets it needs. When you publish, you pick a unique domain and Cursor returns a live link. Source

Two deployment details matter. First, this is a cloud-agent workflow only: Cursor's Apps documentation says the feature is not supported by local agents, so the build happens on Cursor's infrastructure, not your machine. Second, apps can deploy from new or existing repositories on Cursor Origin and GitHub, with support for other git providers like GitLab listed as coming soon. Source

Each app runs in a dedicated hosting account created for your team, or for you on an individual plan, and those accounts are isolated from other teams and users. That is a meaningfully stronger separation story than a shared multi-tenant bucket, though Cursor has not published the underlying architecture in detail. Source

Cursor says its own sales, operations, legal, finance, and engineering teams are already building internal apps this way. The examples it names are unglamorous and telling: event registration and desk-seating tools, account assignment trackers, customer presentations that mix slides and video, onboarding experiences, sales pipeline dashboards, and portals for managing engineering services. These are exactly the long-tail internal tools that never justified a sprint but always slowed someone down. Source

The publish gate

The part of this release most worth studying is what happens before anything goes live. Cursor states that nothing goes live until checks pass and you explicitly publish the app. The changelog describes both deterministic and agentic security checks, aimed at common mistakes like secrets written into app files or exposed credentials, and says a failed check sends the agent back with a list of what to fix. Source

Reporting on Cursor's Apps documentation adds the specifics. Four built-in checks run even if the optional AI security review is switched off: hardcoded secrets in app files block the publish, credential files like private keys or service-account keys block it too, .env files are excluded from the deployed app, and source maps are removed so visitors cannot read the original source. On top of those, an AI security review reads the app's files and blocks serious issues it finds. Source

Team admins get real controls, not just a dashboard. They can see every app and every secret on the team, add their own instructions to the security review prompt, approve publishes the review blocked (with exceptions: malicious code, hardcoded secrets, and exposed credentials cannot be overridden), and unpublish any app. They can also turn Apps on or off for the team and allow or disallow public apps. Notably, admin instructions can add to the review but cannot remove Cursor's built-in checks. Source

Sharing is scoped rather than open by default. You can share an app with specific people, groups, or your whole team, all behind Cursor sign-in. Public sharing exists, but only on individual plans. If a script or bot needs to call your app, you can provision a Cursor API key for it. Source

Pricing and open questions

The changelog says the capability is available on all paid plans with cloud agents enabled. Reporting that checked Cursor's pricing page in November 2026 found that on individual plans, cloud agents start with the 30-dollar-a-month Premium tier; the 8-dollar Plus plan does not include them. Source

The bigger open question is the one the same report flags in its headline: Cursor's documentation, as of November 6, listed no hosting price. The database, the authentication, and the hosting all arrive without a published meter. That does not mean it is free forever; it means the pricing model for the infrastructure half of this product is not yet public, and teams building on it should treat that as an unknown in their cost planning. Source

A second open question is portability. Apps deploy from Cursor Origin or GitHub repositories, and the code stays in a repo you can see, but the hosting, database, and SSO layers are Cursor-managed. Moving a live app off Cursor later means replacing those managed pieces yourself, the same trade-off every managed platform asks you to make. Source

Our read for founders

This section is our assessment, not Cursor's claims. Live Apps is best understood as the editor absorbing the last mile of vibe coding. Until now, the typical vibe coding flow had a cliff at the end: the agent writes the app, then a human figures out hosting, a database, sign-in, and a domain. That cliff is where most internal-tool ideas died, not because they were hard, but because the last mile required a different skill set than the first mile. Cursor is betting that removing it turns occasional experiments into everyday tools.

For non-technical founders, the publish gate is the feature to watch rather than the hosting. The four always-on checks (no hardcoded secrets, no credential files, no .env in the deploy, no source maps) plus an AI review are a reasonable baseline, and they map closely to the first items on any vibe coding security checklist. They do not replace a real review for anything that touches customer data or money. A check that catches an exposed API key does not catch a broken access-control rule, a missing rate limit, or a pricing page that trusts the client.

Our advice is unchanged in shape even if the tools keep improving: use these workflows for internal tools, prototypes, and validation, where the blast radius of a mistake is small. When an app graduates to real users, real payments, or real customer data, it needs the audit, the tests, and the ownership model of production software. That is the point where a vibe coding to production path, or outside help, stops being optional. If you are at that point with an agent-built app, our vibe coding for founders service exists for exactly this handoff.

The competitive frame matters too. Cursor is now competing with Lovable and Replit for the build-and-host journey, not just with other editors for the build step. If this works, expect every AI coding tool to offer managed deployment within a year, and expect the differentiation to move to the things this release only starts on: governance, admin controls, and what happens when a hundred apps built by a hundred people in your company all need to keep working.