# How to put a tool built with AI into production

> A tool built with Claude Code or Cursor works on a laptop. Substrait takes it to production with one command: dev behind sign-in first, then promotion through a security check, with rollback.

URL: https://substrait.build/problems/put-an-ai-built-prototype-into-production/  
Last updated: 2026-11-06

**The question:** Someone on our team built a working tool with Claude Code. How do we get it into production properly?

Run one command, through the AI model that built it. Substrait puts the tool in dev behind your organisation's sign-in, and promotes it to production only through a security check on that exact build.

## Why this is hard

"Just deploy it" sounds like one step.
Underneath it are a dozen specialist steps: packaging the tool, private networking, a secrets vault, least-privilege access, a private path to company systems, a deploy and rollback pipeline, logging, sign-on and data residency.
Each step needs platform and security skills that most builders, and many IT teams, do not have.

## What Substrait does

1. **One command, through the builder's AI model.** The Substrait plugin checks the project, packs the code and tells Substrait to build it.
2. **Dev first, behind sign-in.** Every new tool starts in dev, open only to your organisation. The builder cannot switch that sign-in off.
3. **Going live creates production.** Substrait promotes the build that was tested in dev, without rebuilding it.
4. **The security check runs on that exact build.** A clean build is promoted automatically. A build with too many issues stops for a human approver.
5. **Rollback in one click.** Every release is versioned, and a bad release rolls back to the last good one.

<figure>
  <img src="/screenshots/72_deploy-what-happens@2x.png" alt="Substrait plugin: what happens when a builder runs the deploy command, from checking the project to showing each step as the tool goes live." width="1790" height="470" loading="lazy" />
  <figcaption>What the deploy command does. Product UI. Sample data.</figcaption>
</figure>

## Questions and answers

### What does a tool need before it can go to production?

A place to run, a sign-in, secrets kept out of the code, a security check, a way to roll back, logs, and an owner. Substrait provides all of them as one path, so the builder runs one command.

### Does Substrait rewrite the tool?

No. Substrait runs the unmodified tool the builder produced. It does not rebuild it on another framework.

### Can a production tool be public, for customers?

Yes. In production, sign-in is the owner's choice, because the tool has already passed the security check. A customer-facing form or status page can be public.

## Related

- [One click to live, through the security check](https://substrait.build/capabilities/one-command-go-live/): A builder sends a tool to Substrait with one command, through their AI model, and goes live with one click. Substrait promotes the exact build tested in dev to production, only through the security check.
- [A security check before every release reaches production](https://substrait.build/capabilities/security-check/): Substrait scans every release for leaked secrets, weak code, vulnerable dependencies, container issues and problems in the running tool. A tool reaches production only through a promotion that clears the check.
- [On Substrait's cloud, or inside yours](https://substrait.build/capabilities/your-own-cloud/): Start on a standard setup on Substrait's cloud with nothing to install, or run Substrait in your own cloud account or data centre. In your environment, the whole tool and its data stay inside your perimeter.
- [Nothing breaks quietly](https://substrait.build/capabilities/monitoring-and-alerts/): Substrait watches every tool it runs: crash detection, startup checks and request-level monitoring, with alerts to the builder and the owner about build questions, scan results and release status.
