Vera

Vera is a programming language designed for large language models to write rather than for humans. Its premise is that a model does not need to be right, only checkable, so the language makes correctness something the compiler can verify. Instead of variable names, Vera uses typed positional slot references, which removes a class of naming mistakes that models tend to make. Every function must carry a contract with preconditions, postconditions and declared effects. The verifier uses the Z3 solver to prove what it can at compile time, and the compiler adds runtime checks for most of the rest. A division by a possible zero, for example, is refused with an error and a counterexample instead of failing at run time. Effects are typed and algebraic. Programs compile to WebAssembly, so they can run at the command line, in the browser or as a portable WASI component. The site publishes a SKILL.md and llms.txt for agents that write Vera, along with a FAQ comparing the language with Dafny, Lean and Koka, and a GitHub repository with install instructions. It is intended for developers and researchers exploring model written code, and for teams building agents that generate programs they want checked before they run. The vendor does not publish pricing, and the language is presented as a project to install from GitHub.