Every developer I know who builds AI agents has the same ritual: they stare at LangGraph code, draw mental maps, and pray the agent doesn’t silently freeze mid‑loop. I’ve been that person more times than I can count—including once when I tracked a streaming freeze in Mastra down to a missing synthetic tool‑call chunk.
The problem isn’t that we don’t know how to write agent logic. It’s that we have no simple, visual way to see the agent before it runs.
So I’m building Agent Blueprint.
What is it?
A drag‑and‑drop editor where you connect nodes like LLM Call, Tool Execution, Condition, and Human Approval — and export a working LangGraph Python script with one click.
Why now?
LangGraph usage is exploding, but the developer experience is still raw. We deserve a visual canvas that feels as natural as ComfyUI does for image workflows. I’ve merged 150+ PRs in the frameworks that run modern agents (Mastra, LlamaIndex, LiteLLM), so I’m building this from the inside out — with real understanding of where agents break.
What’s next?
I’m documenting the whole build in public — raw updates, architecture decisions, and a live demo this week. If you’ve ever lost an evening debugging a state graph, follow along. And if you want to influence what it becomes, drop your biggest agent‑debugging pain in the comments.
https://www.linkedin.com/feed/update/urn:li:activity:7467474353769701377/
https://x.com/nehaaaa_6/status/2061705090357215682
Top comments (0)