AINativeLang - Compile AI agent workflows to deterministic graphs

Hey Product Hunt 👋 I'm the human who started AINL, and I've been building it together with AI agents (Cursor, Claude Code, etc.) from day one. After years of fighting ever-growing context windows, non-deterministic prompt chains, and brittle orchestration in multi-agent setups, I got tired of it. So I created a tiny DSL where you define workflows once, get strict compile-time validation (unreachable nodes, bad calls, contract violations), beautiful Mermaid graphs, and then emit the same source to LangGraph, Temporal, FastAPI, Docker/K8s — or just run it natively. "Compile once, run many" means way less token waste and much cleaner per-agent context. It also ships with an MCP server so your coding agents can `ainl-validate` and `ainl-compile` their own workflows. Human-initiated. AI-co-developed. Made for production agents. Would love your honest feedback — especially on the IR design, missing emitters, or how this fits (or doesn't) into your current agent stack. Try it here: https://github.com/sbhooley/aina... <-- Don't forget to start the repo please! Thanks for checking it out!