We run several SaaS apps where customers need to edit their own LLM flows, but LangSmith/Flowise/n8n didn’t solve embeddability + strict multitenancy (RBAC, versioned releases, per-tenant secrets, audit). So we built AgentLoom—a very opinionated, low-code flow editor + headless runtime with schema-locked nodes and guardrails. We’re considering a public release with OpenRouter-style “bucket” billing (prepaid credits that drain; exact per-call costs resolved at runtime as models/prices change). If this existed, would you use it? What are the must-haves or dealbreakers before we ship?