Veritext — AI Document Integrity Guard

Veritext is a lightweight, high-performance integrity guard designed to detect and catch silent AI document corruption during delegated edits (such as multi-agent coding sessions, refactor passes, or code generation). It acts as a strict safety gate between an AI's proposed edits and your version control system.

Key Features

• Rule Guard: Lightning-fast regex checks to catch common lazy AI behaviors (e.g., // ... rest unchanged ...).
• AST Guard: Structural code validation ensuring edits are scoped correctly. Now features full JavaScript and TypeScript syntax validation powered by esbuild.
• Semantic Guard: LLM-assisted verification (via Gemini) to ensure the edit satisfies the prompt without introducing regressions.
• MCP Ready: Built-in Model Context Protocol server for seamless integration with AI agents like Claude Code or Antigravity.

Usage

Installation

git clone https://github.com/your/veritext.git
cd veritext
npm install
npm run build
npm link

Authentication / API Key Configuration

To use the Semantic Guard, you must provide a Gemini API key. While you can use the GEMINI_API_KEY environment variable, GUI Git clients typically do not source terminal environments. To ensure Veritext can always authenticate (especially for Git hooks), you can save your key globally to ~/.veritextrc:

veritext config set GEMINI_API_KEY <your-key>

Command Line Interface

veritext check original.ts edited.ts --prompt "Refactor this code to use async/await"

Git Hook Setup

You can automatically protect your repository by installing a Git pre-commit hook. The hook ensures that every git commit is scanned by Veritext, stopping the commit if any AI-induced code corruption is detected in the staged .py, .js, .ts, .jsx, or .tsx files:

veritext install-hook

MCP Server Integration

Connect Veritext as an MCP server to let AI agents self-verify edits before committing them:

{
  "mcpServers": {
    "veritext": {
      "command": "node",
      "args": ["dist/mcp_server.js"]
    }
  }
}