Vagner Firmino releases V-COS, a governance framework that keeps AI coding agents coherent across development cycles by managing context, hierarchy, and session protocol.

What happened

Vagner Firmino, a solo founder, has released V-COS (Vagner's Cognitive Operating System) as an open-source framework designed to maintain coherence when working with AI coding agents like Claude Code or Cursor. The framework organizes project documentation into three layers—document governance, skills architecture, and agent governance—and can be set up with four core files (AGENTS.md, CLAUDE.md, CYCLE-CURRENT.md, and reference docs). V-COS is extracted from a production SaaS product that has run across more than 25 development cycles.

Why it matters

Solo developers and small teams using AI agents face a recurring problem: without explicit governance, agents lose context between sessions, documents conflict without clear resolution rules, and self-evaluation falls away. V-COS addresses this by making context hierarchy explicit, separating cognitive guidance from technical reference, and embedding mandatory session protocols. The framework is tool-agnostic and sits alongside existing execution frameworks and project methodologies rather than replacing them.

What to watch

The minimum viable setup requires only four files and can be adapted in about an hour using templates from the repository. V-COS is released under the MIT license and includes reference implementations for Claude Code, Cursor, and other tools. The examples/ directory documents how the framework operates in the actual production SaaS product, making it a test case for real-world multi-agent governance.