AI Agents need a new language
Announcing Orchestra Markup Language (“.oml”)
Introduction
Today we’re excited to announce the release of Orchestra Markup Language by Orchestra Labs. Orchestra Markup Language is the result of years of work to create a new dialect for AI and agentic development.
Agents are incredible at writing code — which means everything becomes defined as code. However, traditional code formats leave room for AI bloat and errors.
Existing declarative file formats like .yml and .json are much more succinct, however these present their own challenges for AI:
- A lack of validation of types, formats and allowed values
- A lack of validation for domain-specific rules
- A lack of validation for conditional logic
This creates problems for us as software practitioners:
- Multiple rounds of validation, and in some cases testing, are required to ensure code is correct
- Generating code takes a long time
- The number of tokens used, and therefore cost, is suboptimal
Orchestra Markup Language introduces these validations at the point of write, which means an AI can infer this validation at runtime. Incorrect code is not generated, while developers get granular control over the types of validation that need to be caught.
Orchestra Markup Language includes a lot of Orchestra-specific quirks. A project to extend and abstract these is in active development, and the foundations of .oml will be exposed soon.
It is our hope that as this project develops, we’ll be working on and incubating a new language built specifically for AI Agents.
Read more here.
