Deploy apps from Claude Code, Cursor, Codex, Lovable, Bolt, or v0 with a database, workers, and logs, on NEXUS AI or in your own AWS, Google Cloud, or Azure.
Bring code from any AI tool
Deploy a GitHub repository written with Claude Code, Cursor, Codex, or Copilot, upload a ZIP or folder exported from Lovable, Bolt, or v0, or hand files over from an AI agent through MCP. You can also build the app in the NEXUS AI App Builder, where an agent on Claude, GPT, Gemini, or Grok writes the code, runs checks, and fixes errors before you deploy.
Why AI-generated apps break in production
Code that runs in a chat preview often fails on a server. It listens on localhost instead of 0.0.0.0, hard-codes a port, keeps API keys in the source, stores data in a SQLite file that disappears on redeploy, skips database migrations, or runs Git hook scripts that fail in a headless build. The table below shows how NEXUS AI handles each one.
No Dockerfile required
If the repository has a valid Dockerfile at its root, NEXUS AI uses it. If not, NEXUS AI generates one for Node.js, Next.js, React, Vue, Angular, Vite, Python, FastAPI, Django, PHP, Laravel, Symfony, Ruby on Rails, or a static site. Generated Node.js builds set HUSKY=0 and CI=true, so Git hooks an AI tool added do not break the install.
Deploy the app and its services
The NEXUS AI Docker runtime runs the app with PostgreSQL, MySQL, MongoDB, or Redis sidecars, background workers, persistent volumes, and S3-compatible buckets, and injects the database connection variables, including DATABASE_URL, into the app. Google Cloud Run, AWS App Runner, and Azure Container Apps run single-container apps with supported managed databases.
Run it in your own cloud account
Deploy into your own Google Cloud account on Starter, or AWS, Google Cloud, or Azure on Pro, so the app, its data, and its cloud bill stay in an account you control. NEXUS AI Enterprise runs the whole platform inside your cloud account or on-premises for teams that need AI-generated apps to stay in their own environment.
Operate after launch
Follow deployment status and build logs, open the public HTTPS URL, and inspect runtime logs and health. If a Builder deploy fails, one click sends the build error back to the agent to fix and redeploy. Back up databases on demand or on a schedule, add a custom domain, and roll back a deployment on Pro and higher.
Choose your path to production
Build in AI Builder, then deploy
Best when the app needs several rounds of generation, visual review, and direct code edits.
Create or open a project and select Open AI Builder.
Choose an enabled provider and describe the app. Review Preview and Code; use chat, file edits, screenshots, or click-to-edit to refine it.
Click Deploy in the Builder header, choose an available provider and region, and follow the deployment status to the live URL.
Build a support portal with ticket submission, status tracking, and an admin view. Use Next.js and Prisma. Show loading, empty, and error states. Start with the core app; add authentication in a follow-up.
The instant browser preview does not run a production database. Use the full-stack development sandbox for server behavior, and verify real data after deployment.
After Claude, Cursor, Codex, or another tool writes a repository, deploy that repository from your terminal. Docker sidecars and worker commands apply to the NEXUS AI Docker runtime.
Install the nexus CLI and sign in.
Push the generated, reviewed code to a Git repository. Keep credentials in Secrets Vault, not in the repo URL.
Run the source deploy command, then inspect status and runtime logs.
Replace the repository URL and worker command with your real app. Omit --services and --worker-command if the app does not need them. The process must listen on 0.0.0.0 and the configured port.
Connect an MCP-compatible client to https://mcp.nexusai.run/mcp and approve the needed OAuth scopes. The agent can hand generated files to Builder or deploy a Git repository; these are different tool calls.
For preview, ask the agent to list projects and call nexusai_builder_push with projectId and the complete files array. Open the returned Builder URL and optionally pull later edits with nexusai_builder_pull.
Deploy that Builder snapshot with the Builder Deploy button. Alternatively, put the code in Git and ask the agent to call nexusai_deploy_source.
Have the agent call nexusai_deploy_status and nexusai_deploy_logs, then open the returned URL yourself and verify the app.
Create a support-ticket app in my NEXUS AI project. Push the complete files to AI Builder so I can review the preview. After I approve the code and push it to GitHub, deploy that repository on the Docker provider with PostgreSQL, then show deployment status and the last build logs.
nexusai_builder_push creates a Builder checkpoint, not a production deployment. Ask for deployment only after reviewing the files. Push needs deployments:create; pull and status need deployments:read; logs need deployments:logs.
Best for a focused service or prototype when you want to inspect generated output before a single deploy.
Open a project and expand Advanced: standalone code generator.
Select a configured AI provider and model, enter a specific prompt, then click Generate Code.
Review or edit the generated single file or file set, click Deploy, select a target and required services, then check the Deployments page for status and logs.
Build a Node.js REST API for support tickets with GET /health, create/list/update endpoints, input validation, and clear error responses. Read PORT from the environment and bind to 0.0.0.0.
This generator is separate from the persistent AI Builder conversation. For multi-turn edits, preview, and checkpoints, use AI Builder.
Create a project and generate or import the application files.
Review code and preview behavior, then choose Docker or a supported cloud target.
Deploy from Builder, the Project page, CLI, or an MCP-driven Git workflow.
Verify status, health, logs, public URL, data, and jobs before calling it done.
Deploy code from any AI coding tool
Written with
Get the code out
Deploy it on NEXUS AI
Claude Code, Cursor, or GitHub Copilot
The code is already in your local repository. Commit and push it to GitHub.
nexus deploy source with the repository URL, or Deploy Full App from the repository
OpenAI Codex
Codex cloud opens a pull request in your GitHub repository. Review and merge it.
nexus deploy source, or let Codex call nexusai_deploy_source over MCP
Lovable
Connect the project to GitHub from Lovable.
Import the repository into the App Builder or deploy it with nexus deploy source. Keep the Supabase URL and key as environment variables.
Bolt
Connect to GitHub, or use Export and Download for a ZIP.
Deploy the repository, or upload the ZIP with Deploy Full App
v0
Push to a connected GitHub repository, or Download ZIP.
Deploy the repository, or upload the ZIP with Deploy Full App
ChatGPT or Claude chat
Copy the files, or connect the assistant to the NEXUS AI MCP server.
Upload up to 20 files into the App Builder, or have the assistant call nexusai_builder_push
NEXUS AI App Builder
Nothing to export.
Click Deploy in the Builder header
Why AI-generated apps fail in production, and the fix
Problem
What you see
How NEXUS AI handles it
Listens on localhost
The build succeeds but the URL never responds
Deployment health and runtime logs show the failure. Bind the server to 0.0.0.0 and read the port from the environment.
Hard-coded port
Traffic reaches the wrong port
Set the port in the deploy settings or as a PORT environment variable, and NEXUS AI routes traffic to it.
API keys in the source
Secrets end up in Git history
Store keys in the Secrets Vault, encrypted with AES-256-GCM, and inject them as environment variables at deploy time.
SQLite file for data
Data disappears after a redeploy
Add PostgreSQL, MySQL, or MongoDB with connection variables injected, or attach a persistent volume.
No migrations
Tables are missing on the first request
Run migrations in the start command, for example prisma migrate deploy && npm start, and check the build and runtime logs.
Git hooks in package.json
npm install fails on a husky or prepare script
Generated Node.js Dockerfiles set HUSKY=0 and CI=true.
No Dockerfile
Nothing to build
NEXUS AI generates one, or uses the Dockerfile at the repository root.
Where each plan deploys
Plan
Deployments
Deploy targets
Also included
Free, $0
1, for testing
NEXUS AI
App Builder with 500K AI tokens a month
Starter, $29/month
2
NEXUS AI or your Google Cloud account
Custom domain, 5M AI tokens, 3 team members
Pro, $149/month
5, up to 10 concurrent containers
NEXUS AI or your AWS, Google Cloud, or Azure account
Rollback, 20M AI tokens, 10 team members
Enterprise, custom
Unlimited
Your cloud account or on-premises
SSO and unlimited AI tokens
Frequently asked questions
How do I deploy an AI-generated app in the UI?
Create a project and open AI Builder, generate and review the app, then click Deploy in the Builder header. For a shorter one-shot flow, use Advanced: standalone code generator on the Project page, review the generated files, then click Deploy.
Can I deploy code generated by Claude, Cursor, or Codex?
Yes. Review the files, push them to a Git repository, and deploy the repo through the UI, nexus deploy source, or nexusai_deploy_source. An MCP agent can also push complete files into AI Builder for preview before you deploy.
Can AI-generated apps include databases and workers?
Yes. Docker deployments can include app containers, PostgreSQL, MySQL, MongoDB, Redis, worker sidecars, persistent volumes, S3-compatible buckets, and backup workflows.
Can I preview an AI-generated app before deploying it?
Yes. Generate or upload files in AI Builder, or use nexusai_builder_push from an MCP client. Review the Preview and Code tabs before deploying. Browser preview is not the production database; test full-stack behavior in the development sandbox and after deployment.
What is the difference between AI Builder and the Project page code generator?
AI Builder is a persistent, multi-turn workspace with preview, direct code editing, checkpoints, and deployment. The Project page also has an Advanced: standalone code generator for a focused prompt-to-code-to-deploy flow.
Does nexusai_builder_push deploy my app?
No. It saves the complete file set as an AI Builder checkpoint and returns a preview URL. Deploy from the Builder after review, or put the code in Git and use nexusai_deploy_source for an agent-driven source deployment.
Can the CLI deploy a local folder directly?
The documented nexus deploy source example builds from a Git repository. In the dashboard, Deploy Full App can accept a local source folder or ZIP. If you edit locally with an AI coding tool, push the reviewed code to Git before running the CLI source command.
How do I verify the deployment?
Use the Builder status banner, Deployments page, nexus deploy status and nexus deploy logs, or the MCP status and logs tools. Open the public URL and test routes, database access, jobs, uploads, and health before changing production traffic.
How do I deploy a Codex-generated app?
Merge the pull request Codex opened in your GitHub repository, then run nexus deploy source with the repository URL, or deploy the repository from the dashboard. If Codex is connected to the NEXUS AI MCP server, it can call nexusai_deploy_source itself and read the status and logs back.
How do I deploy an app from Lovable, Bolt, or v0?
Connect the project to GitHub from the tool, then deploy the repository with nexus deploy source or import it into the AI App Builder. Bolt and v0 can also export a ZIP, which Deploy Full App accepts. Lovable apps that use Supabase need the Supabase URL and key set as environment variables.
Do I need a Dockerfile?
No. NEXUS AI uses a valid Dockerfile at the repository root when there is one, and otherwise generates one for Node.js, Next.js, React, Vue, Angular, Vite, Python, FastAPI, Django, PHP, Laravel, Symfony, Ruby on Rails, or a static site.
Can I deploy an AI-generated app into my own AWS account or VPC?
NEXUS AI deploys into your own Google Cloud account on Starter and your own AWS, Google Cloud, or Azure account on Pro, using Cloud Run, App Runner, or Container Apps in the region you choose. For a platform that runs entirely inside your own network, NEXUS AI Enterprise deploys in your cloud account or on-premises.
What happens if the deploy fails?
The build and runtime logs show the error in the dashboard, the CLI, and the MCP logs tool. For apps built in the AI App Builder, one click sends the failure back to the agent, which fixes the code and redeploys.
Is it free to deploy an AI-generated app?
Yes. The Free plan includes one deployment on NEXUS AI for testing and the AI App Builder with 500K AI tokens a month. Starter ($29 a month) adds a second deployment, a custom domain, and Google Cloud Run. Pro ($149 a month) adds AWS, Google Cloud, and Azure.