How to deploy code generated by an AI
Getting a working app out of an AI is something anyone can do today: you describe an idea, you get the code for a site, a bot, or a service, you run it on your machine — and it works. Then everything runs into the part the AI usually covers in a single line, "deploy it to any host": the server, the domain, the environment variables, HTTPS. For someone who did not write the code by hand, that is a wall.
Let us walk through how to take generated code all the way to a live link without digging into how servers work. The approach is the same no matter which tool wrote the code — ChatGPT, Claude, Cursor, or anything else: what matters is not the generator, but that you have a project folder that runs.
Ask the AI to collect the project into one folder
AIs often hand you code in pieces across a chat. Ask it to gather everything into a single project with a clear structure and one command to start it, and to produce a list of dependencies along the way: requirements.txt for Python or package.json for Node.js. That gives the platform something to build the environment from, so you do not have to install libraries by hand.
Ask the AI where the secrets live in the code
AIs often paste API keys and tokens straight into the code "as an example". Before you put it online, ask it to move all such values into environment variables and keep them out of the project files. That is both safer and needed to run: when you set up the project, the platform will ask for those values separately.
Check that the port comes from the environment
Generated code often listens on a hard-coded port like 3000 or 8000. Ask the AI to read the port from an environment variable (usually PORT). Then the platform sets the right port itself, and the app becomes reachable at a link without any manual network setup — this is the single most common reason for "works locally, but not on the host".
Run the project through the free check
Before you sign up, you can upload the project to the free Netrun check without logging in: it recognizes the language and the type of project and tells you whether it will run as is or something is missing. It is a handy way to catch the rough edges of generated code early, without wading through server logs.
Upload the code to Netrun
Upload the project as a ZIP archive or import it from a GitHub repository, if the AI tool already pushed the code there. Netrun figures out the language on its own — Python, Node.js, Go, and others — and builds the app. You do not have to pick a server type, a language version, or a build method by hand.
Answer the questions and get your link
The platform asks for the values of the secrets you moved out of the code, builds the project, and puts it online at an address with HTTPS. Build logs are visible in real time, and if something is missing, the reason is shown in plain language — which helps when you are not the one who wrote the code.
AI-generated code stops being just a folder on your computer the moment it gets a link. Netrun takes on exactly the part the AI usually skips — the server, the build, HTTPS, and restarts. Upload your code to Netrun, or first run it through the free project check.
Common questions
Can I put online code that ChatGPT or Claude wrote?
Yes. It makes no difference to the platform who wrote the code — a person or an AI. What matters is that the project collects into one folder, runs with a single command, and takes its port from an environment variable. Netrun detects the language and builds the project itself.
What do I do if the app works locally but not on the host?
The reason is most often a hard-coded port or keys baked straight into the code. Ask the AI to read the port from the PORT environment variable and move the secrets outside. The free project check in Netrun surfaces these issues in advance.
Do I need to understand the code to deploy it?
You do not need a deep understanding. It is enough to collect the project into one folder and move the secrets out. From there Netrun builds and runs the app itself, and shows errors as plain text rather than a stack trace.
Does this work for code from Cursor, v0, or other AI tools?
Yes. The approach is the same for any tool: if you have a project that runs locally, you can upload it as an archive or import it from GitHub, and Netrun will deploy it.
Is it safe to upload generated code that has keys in it?
Keys and tokens should not sit in the code. Move them into environment variables, and Netrun will ask for those values separately during setup — there will be no secrets in the archive or the repository, and in the platform database they are stored encrypted.
Which languages and technologies are supported?
Python, Node.js, Go, Rust, Ruby, PHP, Java, .NET, Deno, Bun, Elixir, static sites and bash scripts. You can bring your own Dockerfile or docker-compose, but more often the stack is detected from your code automatically.