VVM.run is an open-source programming language and framework designed for intelligent runtimes, specifically to structure complex workflows for large language models (LLMs) with tool access, treating them as general-purpose "agentic computers". It allows users to define declarative workflows in .vvm files, specifying workers (agents) with roles, models, prompts, skills, permissions, dependencies, parallel execution, retries, and quality gates (e.g., requiring citations, no hallucinations, or specific comprehension standards). Execution relies on Claude Code (or compatible interpreters) to parse the file, spawn sub-agents, manage context, enforce constraints, and produce reliable outputs. This bridges natural language prompting and traditional programming, making agent behaviors more modular, reusable, and robust for tasks beyond simple chats, such as research, analysis, or multi-step automation.

The project launched in January 2026 with community funding support via the $VVM token on Solana (via BagsApp launchpad), and emphasizes semantic understanding over regex-based routing.