BAML

BAML is a programming language built specifically for developing AI agents. It draws inspiration from TypeScript's syntax and design philosophy but targets a new paradigm: agentic coding. The language aims to prevent context pollution and churn when coding with AI, applying type safety and structured error handling to the unpredictable world of LLM-powered agents.

BAML's design principles include making undesired states unrepresentable, avoiding viral edits, and making nondeterminism observable and testable. It offers tools tailored for agent development rather than general-purpose IDEs. The language feels like TypeScript but with better error handling and no 'any' type — a deliberate trade-off between dynamism and runtime safety. The project draws an analogy to how Assembly matched hardware, Java matched operating systems, and JavaScript matched the web — BAML is positioned as the language for the agentic coding era.

The primary audience is developers building AI agents who want a language that reduces bugs and cognitive overhead. By enforcing type safety at compile time and providing observability into agent behavior, BAML helps engineers create more reliable LLM-driven applications. It's particularly useful for teams that find existing general-purpose languages (like Python or TypeScript) lacking in guardrails for agentic workflows.