Our story

We stopped trusting green builds. So we built a way to prove them.

Passing tests and high coverage still let bugs reach customers. OrangePro checks whether your tests would actually fail if the code that just changed were wrong, and leaves the proof in your repository.

Aamir Siddiqui, founder 5 min read

The night the dashboard lied

The release was ready. QA signed off. The dashboard glowed green. Everyone went home.

At 6:12 a.m., Slack exploded. Carts were failing in some regions. A discount rule clashed with a new pricing service. No crashes, no red alerts, just a silent revenue leak and a flood of support tickets.

Every test had passed. None of them would have failed if that pricing logic was wrong. That gap, between tests that pass and tests that protect, is what OrangePro is about.

Twenty years of green builds

I have spent about 20 years leading QA and engineering at Oracle, Ford, Salesforce and ServiceNow. We obsessed over code coverage and fast pipelines. We still could not answer the question the business cared about: if this change were wrong, would anything catch it?

Coverage tells you which lines ran. It does not tell you whether a single assertion would fail if the behavior on those lines broke. Teams learn the difference in production.

What changed

Our earlier platform read stories, tickets and incidents and proposed test cases from them. It helped teams see gaps, and it taught us a lot. Then AI agents started writing a large share of the code.

Writing code got cheap. Reviewing it did not. When an agent writes the code and the tests for it in the same pass, a green build says even less than it used to. The bottleneck moved from writing code to proving it works.

So we rebuilt OrangePro around one question that does not depend on who or what wrote the code: would your tests fail if the code that just changed were wrong?

What OrangePro does now

After merge and before release, OrangePro maps the functions a change touched and ranks them by risk and by how much real test evidence protects them. For the risky ones, it breaks the behavior on purpose in an isolated copy and reruns your tests.

Where nothing notices, it can draft a focused test with the model provider you choose. A test counts only if it passes on your code and fails on the broken copy. You get a test-only patch with that pass and fail receipt for every test. The tests live in your repository and run in your own CI.

It is not a pull request check and not a release gate. It does not replace CI, code review or end-to-end testing. If it cannot run your tests, it says so, names the blocker and calls the result unknown.

Why the engine is open source

Proof you cannot inspect is just another claim. The engine, @orangepro/mcp-server, is MIT licensed and runs on your machine. Mapping and ranking need no model, and your source does not leave your environment. Models and agents will keep changing. The proof should stay with you.

Who we are

OrangePro Technologies Inc. is a Delaware corporation, backed by Plug and Play. We want release days to be boring again.

AI made writing code cheap. Proving it works is the new bottleneck.

Want more boring release days?

Run the free report on a public repository, or bring one real pull request and we will run OrangePro on it with you.

Get the free report