Jonathan Miranda

Compyle - The coding agent that asks before it builds

Stop wasting time debugging autonomous agents that built the wrong thing. Compyle keeps you in the driver's seat - asking smart questions to understand requirements, creating planning artifacts before coding, and checking in when decisions need to be made. Now with GLM 4.6, Claude Sonnet 4.5, and Claude Haiku 4.5 support. Get 20% off all plans today with code PH20!

Add a comment

Replies

Best
Jonathan Miranda

Hey Product Hunt! 👋

I'm Jonathan, co-founder of Compyle, and we're thrilled to launch today with an exclusive 20% discount on all plans - just use code PH20 at checkout (today only)!

The Problem We Solved 🎯

We built Compyle because we were tired of "autonomous" coding agents that:

- Disappear for 20 minutes and come back with the wrong solution

- Write 2,000 lines of code you don't understand

- Make countless assumptions without asking

- Force you to debug and refactor their decisions

Our Different Approach 🤝

While other agents chase maximum autonomy, we believe the future is collaborative AI. Compyle works WITH you, not for you:

Asks first, builds second - No more unwanted surprises

You make decisions, we handle execution - Stay in control

Understand everything - Because you guided every decision

Planning artifacts - See the roadmap before a single line is written

Proactive collaboration - We ask questions so you don't have to prompt-engineer fixes

What's New 🚀

- Multi-model support: Choose between GLM 4.6, Claude Sonnet 4.5, and Claude Haiku 4.5

- Smarter question flows: Only asks what matters

- Better at complex, open-ended work: From new features to entire products

Who It's For 💡

- Developers tired of black-box code generation

- Founders who need to understand what they're building

- Teams who value maintainable, debuggable code

- Anyone who's ever spent an hour fixing what an AI built in 5 minutes

Special Launch Offer 🎉

Today only: Get 20% off any plan with code PH20 at checkout!

Mark (my co-founder) and I are here all day to answer questions. We'd love to hear about your coding agent frustrations and how we can help!

What features would you want to see next? Let us know! 👇

Mark Nazzaro

Let's goooo. Using Compyle to make Compyle

Jonathan Miranda

@mark_nazzaro Compyling the future one task at a time

Van de Vouchy

Hey Jonathan, quick one how steep is the learning curve for someone coming from classic fire-and-forget AI agents?

Piyush Mudgal

@vouchy Hey! I think Compyle’s really easy to pick up. The only part that might take a bit of getting used to is how it works with GitHub branches and pull requests.

Jonathan Miranda

@vouchy +1 Piyush's reply - it's really easy to get up and running with Compyle. Let us know if you try it!

Gerome Elassaad

As a software engineer, I genuinely love this tool. The planning workflow and the questions it surfaces are spot-on. Any time I was stuck tracking down why a build was failing, it analyzed the issue and helped me resolve it quickly. Really well done. Excited to see where you take this next—congrats on the launch!

Jonathan Miranda

@gerome_elassaad Thank you for the support Gerome! A lot of cool stuff coming soon :)

Piyush Mudgal


Hey, I really like this project.
It honestly feels like working with a skilled teammate who’s right there with me.

The agent asks thoughtful questions, confirms details before acting, and helps me refine my ideas as I build — which I really enjoy.

Its task-based workflow — where each task creates its own branch and PR — makes collaboration effortless and keeps my Git history much cleaner.

I also appreciate how responsive the team is and how quickly they address any issues.

The UI is sleek, the automation feels reliable, and I’m excited for future updates like live preview and in-app code editing. Overall, a solid product with a ton of potential.

Jonathan Miranda

@pyasma Thank you for the support, Piyush! We're excited for the previews and editor too 🚀

Cruise Chen

This proactive approach seems impressive - how many rounds of questions could Compyle collect before finalize? Congrats @jonathan_miranda1 !

Jonathan Miranda

@cruise_chen Thank you! It can ask quite a bit, and we're working on 10-100x ing the amount of questions. Also - it doesn't just ask questions at the beginning/planning phase, we have an orchestration of background agents that ask questions throughout the whole task.

Chilarai M

This is one of the best ideas I have ever seen for building apps with AI. Yes, the issue is that we are losing control, and we need to regain it in order to create meaningful apps with AI.
Good job! Will try this tool

Jonathan Miranda

@chilarai Appreciate the support Chilarai!! Let us know how it goes. We're here if you need anything

Ciprian Balanica
This actually sounds like it might solve a lot of the current issues agents have. Though I'm wondering, Claude Code and recently OpenAI's Codex already do similar things, where they ask questions for stuff that wasn't explained properly and also write documentation on the side. It seems like your tool does pretty much the same thing, maybe with a couple more questions, but what other differentiations would there be between your product and the aforementioned products? Great idea start though. Also, do you plan on integrating somehow with IDEs like VS Code?
Jonathan Miranda

@nair0 Thanks, Ciprian! We're actually quite different. All the other coding agents try to collect just enough information at the beginning - and then run off into 10+ minute, fully autonomous sessions. They never proactively check in with you, they code code code and leave it up to you to force check-ins. Compyle completely flips that by working with you throughout the whole task lifecycle.

@jonathan_miranda1 This addresses a real pain point in AI coding—the black box problem. The vision of keeping developers "in the driver's seat" with question-driven collaboration rather than autonomous chaos is exactly what the industry needs. Avoiding those wasteful 20-minute builds that do the wrong thing is a game-changer for productivity. The planning artifacts approach with clarifying questions shows maturity in understanding how complex, open-ended work actually happens. This could fundamentally shift how teams integrate AI into their development workflow—not replacing judgment, but augmenting it intelligently.

Jonathan Miranda

@kjosephabraham Thank you for your support, Joseph!

The decision-making power should be in the developer's hands, not AI's. Coding agents should only be handling the mechanical execution of code and driving the conversation forward, if need be, but never making decisions on our behalf.

Mykyta Semenov 🇺🇦🇳🇱

Great idea. Is it suitable for refactoring old legacy projects?