Guides
Deploy internal tools built with AI: a 3-step guide
Deploy an internal tool you built with Claude, ChatGPT or Cursor in 3 steps: connect Inhaus, ask your AI tool to put it live, then share it with your team.
You can deploy an internal tool built with AI in three steps: add the Inhaus connector to your AI tool, ask it to put the tool live, then ask it to share the tool with your team. The tool goes live behind your company login, so only the people you choose can open it.
It works with any AI tool that supports MCP connectors, including Claude, ChatGPT, Cursor, Claude Code, VS Code and Windsurf. You don't need a hosting account or build settings.
Before you start
- A tool your AI built: a calculator, tracker, checklist, dashboard or form. It should be a web app in HTML, JavaScript or TypeScript, with or without server code.
- An Inhaus account and a workspace. Create one at inhaus.dev/getstarted. Pick Company if you want to share with people at your company behind your company login. Your first 7 days are free.
The 3 steps
- Connect your AI tool to Inhaus
- Ask it to put the tool live
In the chat or project where the tool is, ask:
Deploy this to Inhaus as "Commission Calculator".Your AI tool sends every file to Inhaus. Inhaus installs packages, bundles the code, runs a security scan and puts it live. Usually in under a minute, you get back the link (
https://commission-calculator--acme.inhaus.dev), the version number and the scan result. Only you can open it so far. - Share it with your team
Share it with priya@acme.com and the sales team, view only.Inhaus emails each person the link. They sign in with their company Google or Microsoft account and use the tool. They don't need an Inhaus account.
That's it. The rest of this page covers what happens next.
What happens on each deploy
- Queued. Inhaus stores the files and creates a new version. Large apps are zipped and uploaded first, on their own.
- Building. Packages from
package.jsonare installed and TypeScript and JSX are bundled. - Scanning. The security scan checks for keys in code, calls to outside websites and known vulnerable packages.
- Live. The version goes live behind the Inhaus gateway.
If the build fails, the previous version stays live, and your AI tool gets the error so it can fix it and try again. See Deploy an app.
Keep it working
- Change it. Ask for the change and say "redeploy". It becomes a new version at the same link. Everyone keeps their access.
- Undo a change. "Roll back the commission calculator to the old version." It's live again in seconds.
- Fix errors. "Check the logs for the commission calculator and fix the errors." Logs are kept for 7 days. See Logs.
- See who uses it. "Who opened the commission calculator this week?" See Access log.
FAQ
Which AI tools can deploy to Inhaus?
Any tool that supports remote MCP connectors with sign-in, including Claude, ChatGPT, Cursor, Claude Code, VS Code and Windsurf. See Connect your AI tool.
What kinds of tools can I deploy?
Web apps written in HTML, JavaScript or TypeScript: static sites, including React, and apps with server code. Python is planned. See Deploy an app.
Do I need to know how to code?
No. If you can ask an AI tool to build something, you can ask it to deploy and share it. Setup is pasting one URL.
What if the security scan fails?
The tool still goes live for you and the people you already added. It just can't be listed for the whole company until the blocking issues are fixed. Ask your AI tool: "Fix the security issues Inhaus found and redeploy." See Fixing issues.
How long does a deploy take?
Usually under a minute. If a large build takes longer than 90 seconds, it keeps going and goes live on its own.
What happens after the 7 free days?
Your tools keep running and people keep access. Deploying and sharing need a plan. See Plans.
Next steps
- Quickstart: the same flow, with the setup for each AI tool.
- How to deploy vibe-coded internal tools securely
- Share an app