Context
Multi-agent workflows are difficult to reason about as code and worse to hand over. The team that inherits one inherits a pipeline they cannot see, and the first thing they do is rebuild their own mental model of it from the source, slowly and incorrectly.
Built
A visual, AI-native workflow builder for composing and running multi-agent systems.
It is also the proving ground where the LOGIC.md reasoning-contract specification was first validated at scale, which matters more than it sounds: a specification nobody has run in anger is a document, not a standard.
Stack
TypeScript, Next.js, React and Node.js, with a multi-provider LLM abstraction and MCP.
Outcome
In active development as the flagship product build. The specification work it validated is now published and independently used, which is the useful half of the result: the product proved the standard, and the standard outlived the product cycle.