Launching today

Regent

Launching today

Know when your AI changes behavior

8 followers

Know when your AI changes behavior

8 followers

Visit website

Reliability is the next frontier for AI apps. We were already shipping MVPs. It's time to turn them into products. Existing observability tools do for LLM apps what logs did for conventional ones: they tell you what happened, not what changed. Regent is the first regression testing layer for agentic apps. It can run semantic diffs on your agent's entire execution trace for any critical inputs, all before you merge the PR. Posts the results right in your Github. No more finding out from users.

Free

Launch tags:SaaS•Developer Tools•Artificial Intelligence

Launch Team / Built With

Wispr Flow: Dictation That Works Everywhere — Stop typing. Start speaking. 4x faster.

Stop typing. Start speaking. 4x faster.

Promoted

Maker

📌

Hey Product Hunt 👋 I'm Agrim, and this one's been a while coming. It started with Jarvis, my AI secretary. Every time I changed a prompt or tweaked the architecture, which was constantly, I had no idea what I'd quietly broken. So I'd manually test scenario after scenario, every single time, knowing it would only get worse as the app grew. I looked for something that could just tell me "this behaviour changed." Nothing existed for agentic apps. So I built Regent. The thing I'm most proud of honestly is the onboarding. I was obsessive about it. If integrating a testing tool feels like a project in itself, nobody does it. So I kept going until it was 2 lines of code. That part took some time, but definitely worth it. What Regent does is simple in concept but surprisingly hard in practice: it records how your agent actually behaves, including traces, nested calls, everything. And diffs it against that baseline on every PR, posting the results straight into your GitHub comment. Not just text output, the full execution trace, actions, flow, decisions. If you're building LLM apps and you've felt this pain, I'd genuinely love to hear from you. Honest first impressions, what resonates, what feels off, all welcome.

Report

14h ago

Yeah feels like there’s a gap here tbh. Right now we’re mostly just manually testing / eyeballing outputs, which isn’t ideal. Let’s try plugging this into our workflow and see how it goes.

Report

2h ago

Maker

@sresth_anand Yes! This idea was actually born out of neccessity, we actually ran into this problem ourselves while working on our own agentic application. Looking forward to having you.

Report

1h ago

Apple Health Wrapped 2025

Super interesting problem statement. Love to see how the product evolves to solve it.

Report

1h ago

Maker

@nihal_goyal Agreed Nihal! Lots to come here, we’re excited!

Report

33m ago

Founder Mode Checker

I tried it yesterday and it works great! Really like the product. I would probably start using this in my main workflow now.

Report

2h ago

Maker

@theishangoswami Thanks for the feedback! Glad it actually helped. We have a lot more interesting things in the pipeline, excited to get them out!

Report

2h ago