Cursor has turned its coding agent into a deployment tool: developers can now describe an app in plain language and get back a live, shareable web app. A cloud agent scaffolds the app, spins up a Postgres database when needed, sets up authentication behind Cursor’s sign-in, and provisions hosting. The capability is live now on all paid plans with cloud agents enabled, and Cursor says its own teams already use it for internal tools.
## How it works
The flow starts with a cloud agent and an explicit request to build an app. Cursor then handles the parts that keep a project on localhost: scaffolding, database setup, authentication and hosting. When the app is ready to publish, the developer picks a unique domain and Cursor returns a live link, shared with chosen people, groups or the whole team.
Cursor points to its own use as proof the feature is ready for real work. Staff across sales, operations, legal, finance and engineering built event registration tools, desk seating planners, customer presentations, onboarding experiences, sales dashboards and portals for managing engineering services. Apps can deploy from new or existing repositories on Cursor Origin and GitHub, with GitLab support promised soon.
## Guardrails before anything goes live
Publishing runs through deterministic and agentic security checks that look for leaked secrets and exposed credentials. If a check fails, the agent explains what needs fixing, and nothing goes live until the checks pass and the developer explicitly publishes. Team admins can see every app and secret, add their own review rules, approve blocked publishes and unpublish apps.
Individual plans can also share apps publicly, while team apps stay behind Cursor sign-in. Developers can provision a Cursor API key so a script or an automated bot can use an app, and Cursor says the same changelog drop also makes its Agent SDK generally available, for defining agents as code and deploying them to Cursor or elsewhere.
## Our read
Live Apps closes the last gap in Cursor’s pitch: it is no longer just where code gets written, but where the finished thing runs. That matters because the hardest part of internal tools was never writing them, it was deployment, databases and sign-in, the chores Cursor now absorbs.
The question is whether teams trust an agent-built app with real data, and Cursor’s answer is process rather than promises: security checks before publishing, admin visibility over every app and secret, and an explicit publish step. Anthropic has turned Claude into an [enterprise app store](https://geroundtable.com/claude-enterprise-app-store/) of its own, and the race is now over who runs the software agents build.
## Why It Matters
If Live Apps works as described, the distance from idea to working internal tool collapses to a single prompt, and the people who feel it first will be non-engineers in sales, operations and legal teams, who have waited longest for engineering time.
For Cursor, the stakes are bigger than a feature: every app it hosts deepens its hold on a team’s workflow, at a moment when [coding agents](https://geroundtable.com/cognition-swe-2-flash-everyday-coding/) from every lab are competing to become the place where software gets built and run.
## Background
Cursor, built by Anysphere, has grown from an AI-first code editor into a cloud-agent platform, with its agent tools now used across software teams. The company spent 2026 adding cloud agents, remote control from its iOS app and a marketplace for its Origin platform, and Live Apps is the step that lets those agents ship finished products, not just code.
Source: [Cursor’s changelog](https://cursor.com/changelog) and [AlphaSignal’s report](https://alphasignal.ai/news/cursor-s-build-live-apps-turns-a-text-prompt-into-a-hosted-web-app)
0 Comments