LMOS ADL

A structured, model-neutral language and toolkit for defining and deploying reliable AI agents.

For business experts, enabled by engineers.

Why ADL

ADL targets enterprise and business‑process complexity.

Lets business author flows & iterates quickly.engineers wire tools, policies, and data
Engineers wire tools, policies, and data
Replaces fragile prompting with governed, versionable Agent Business definitions
Built‑in validation, fallbacks, and recovery for reliable operations

What is ADL?

ADL is a declarative, model-agnostic language for defining agent behavior. It separates business logic from prompts, making agents easier to build, change, and trust.

Think of it as the SQL of agent behavior: a shared language that standardizes how agents think, act, and recover.

Read the docs

Set up in your organization

Start with a simple Spring Boot app. Engineers configure once; business teams iterate on ADL.

For engineers

Bootstrap with the Arc Spring Starter
Wire tools/APIs, policies, and data sources
Expose ADL Playground at /chat/index.html

For business teams

Access ADL Playground exposed by engineering team
Create ADL's for your agent usecases
Define steps, solutions, mock tools and fallbacks — no code
Test in Chat and iterate quickly

Note: Engineers can wire APIs or system tools; business users can mock tools. Once initialized, business teams own their use cases.

ADL Starter for Engineers

Try it now

Open the Playground to test ADL interactively. Create a UseCase, validate responses, and iterate fast.

Open Playground