The Problem
Enterprise teams want to build with AI agents but the tooling forces a choice: either write everything in code (which locks out non-engineers and slows iteration) or use a hosted no-code tool (which forces architectural decisions you didn't sign up for). There was no middle ground for teams that wanted real control over multi-step agent workflows without writing orchestration code.
The Approach
We built Modular9 — a visual multi-agent workflow builder that lets teams design, connect, and deploy agent pipelines through a drag-and-drop canvas interface. Built on Next.js and React Flow, with a node graph that maps directly to the underlying agent runtime.
Each node is a real agent or tool — not a wrapper around an opaque service. Connections describe data flow and control flow. The canvas reflects the actual execution graph, so what you see is what runs. Non-engineers can compose workflows; engineers can extend them with custom nodes.
The Outcome
Modular9 makes complex agent orchestration accessible to teams that previously couldn't build it at all. Workflows that would have required a sprint of engineering work are composed in an afternoon. The visual layer becomes the documentation — there's no gap between the diagram and the running system.
It's the kind of tool we built because we needed it ourselves. Now other teams use it the same way.