HILO — Agentic DSL for GPTs, AIs, and Operational Tooling

HILO is a compact, agent-focused DSL crafted so GPT-class models, automation AIs, and operational tooling can parse, reason about, and coordinate complex workflows. It stays human-readable while foregrounding agent roles, policies, tools, and orchestration primitives—turning .hilo files into executable playbooks for autonomous systems.

Status: Draft v1.0 • 2025-10-17

Why HILO?

• Simple, explicit syntax: braces for blocks; newlines end statements (semicolons optional).
• Gradual types: types help structure; omit when obvious.
• Functional core, OO shell: pure functions + lightweight classes/records/enums.
• First-class workflows: agent, task, and workflow primitives.
• Safety & governance built in: policy blocks, explicit tool declarations, and redaction hints.
• Parse-friendly: reserved keywords and EBNF are provided; tokens avoid ambiguity.

HILO is not a runtime. It is the coordination DSL that glues GPTs, AIs, and software tooling together—codifying how they collaborate safely and correctly.

Quick look

module demo.hello

func greet(name: String = "world") -> String {
  return "Hello, " + name + "!"
}

task main() {
  let msg = greet()
  io.print(msg)
}

Getting started

1. Add bin to your PATH or invoke it via ./bin/hilo.
2. Run hilo init <target-directory> to scaffold a fresh project; the script copies the language spec, sample agents, and bootstrap guidance into the target.
3. Open the entry AGENTS.hilo and follow its instructions, then explore imported .hilo files as needed.

HILO is designed to coordinate systems. When you implement it, treat HILO as a source of truth for intent, contracts, and workflows—even if your runtime is Python, JS/TS, Go, Rust, or another stack.