xiaolai/vmark

What it solves

VMark is a local-first, plain-text workspace designed for seamless collaboration between humans and AI. It eliminates the translation layer between how users and AI agents read and write files, allowing both to work directly on the same artifacts (such as Markdown, JSON, YAML, and code) in a format-aware environment.

How it works

VMark provides a multi-format editor that renders different file types based on their schema. For example, GitHub Actions workflows are displayed as workflow graphs and dependency files like package.json are shown as dependency trees. It integrates natively with the Model Context Protocol (MCP) to connect with AI assistants like Claude Desktop and Gemini CLI, enabling AI "Genies" for inline writing assistance while maintaining a local-first approach where documents stay on the machine.

Who it’s for

It is designed for developers and writers who use AI agents to help manage their projects, write documentation, and edit structured data files without needing to switch between a generic text editor and an AI interface.

Highlights

• Schema-Aware Previews: Renders specific views for known artifacts like Cargo.toml or GitHub Actions workflows instead of generic trees.
• AI-Native Integration: Native MCP support for Claude Desktop, Claude Code, and Gemini CLI.
• Multi-Format Support: Handles Markdown (WYSIWYG and source), JSON, YAML, TOML, Mermaid, SVG, and HTML.
• Local-First: No cloud accounts or analytics; all documents remain local.
• CJK Optimization: Includes over 20 formatting rules specifically for Chinese, Japanese, and Korean text.