Launching today

AgentGraph Studio

Launching today

Preflight AI workflows before you run them

1 follower

Preflight AI workflows before you run them

1 follower

Visit website

AI Infrastructure Tools

•

AI Code Testing

AgentGraph Studio is a portable engineering tool for AI workflows. Today, it imports CrewAI Python or AgentGraph JSON, turns them into deterministic Preflight evidence, lets you improve the design, re-verify it, and export artifacts. It analyzes imported Python without executing it and never presents static evidence as runtime truth. Built with GPT-6 Astra, it is evolving toward a verification layer connecting expectations, static evidence, and observed runtime behavior—under human control.

Free

Launch tags:Developer Tools•Artificial Intelligence•OpenAI Day

Launch Team / Built With

OpenAI

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

Astra changed the ambition of AgentGraph Studio by making it practical for me to work on the product as a coherent engineering system, not just isolated features. I used GPT-6 Astra as the default executor for complex, multi-step repository work across implementation, documentation, verification, and follow-through. That let me push further on the product’s real goal: helping people understand, evaluate, improve, verify, and own increasingly complex AI workflows. Just as importantly, Astra did not replace the product’s deterministic core. AgentGraph Studio keeps static facts deterministic, separates static evidence from runtime truth, and keeps semantic changes under human control. Astra helped me move faster across the engineering surface while preserving those boundaries. The result is a more ambitious direction: not only a visual workflow builder, but a portable engineering and verification layer for AI workflows as AI makes building them dramatically easier.

Report

10h ago

Maker

📌

Hi Product Hunt 👋 I started building AgentGraph Studio because creating AI workflows is getting easier much faster than understanding and verifying them. Models like GPT-6 Astra can now carry complex software engineering work across many steps. That’s exciting, but it also creates a new problem: when AI can produce increasingly large systems, how do humans understand what was built, distinguish evidence from assumptions, improve it safely, and keep ownership of the result? Today, AgentGraph Studio lets you import supported CrewAI Python or AgentGraph JSON, inspect deterministic static Preflight evidence, improve the workflow manually, verify it again, and export portable artifacts you own. Imported Python is analyzed rather than executed just to inspect it, and static analysis is never presented as observed runtime behavior. GPT-6 Astra has also been part of how I build AgentGraph Studio itself, helping with complex, multi-step engineering work while the project keeps deterministic checks, explicit evidence, and human-controlled changes at the center. Where I want to take it next is bigger than static Preflight. The goal is a portable engineering and verification layer for AI workflows: connect configured expectations, static evidence, and eventually observed runtime evidence; evaluate architectures against scenarios and acceptance criteria; propose improvements without silently changing semantics; and let users verify and own the systems AI helps them create. The product direction is: Understand → Evaluate → Improve → Verify → Own. As AI makes building dramatically easier, I think the harder problem shifts from “Can we generate this?” to “Can we understand, trust, improve, verify, and own what was generated?” That’s the problem AgentGraph Studio is trying to solve. I’d love to hear what kinds of AI workflows you’d want to inspect or verify with it.

Report

7d ago