VibiumDev/vibium

What it solves

Vibium acts as a verification and automation layer for AI coding agents. It solves the problem of agents needing reliable, standardized ways to interact with web browsers to perform tasks, verify claims, and check their own work without relying on proprietary or complex automation protocols.

How it works

Vibium provides a browser automation toolkit that agents can use as a "skill." It uses the WebDriver BiDi standard to communicate with browsers, allowing agents to navigate pages, fill forms, click elements, and take screenshots using simple CLI commands or API calls. It can be integrated as a standalone CLI tool, an MCP (Model Context Protocol) server for chat interfaces, or as a library within JavaScript, Python, or Java applications.

Who it’s for

It is designed for developers building AI agents who need to give their models the ability to browse the web, verify information, or automate browser-based workflows. It is suitable for those looking for a lightweight, cross-platform solution that handles browser setup automatically.

Highlights

• AI-native: Designed specifically for agents to learn browser interaction as a skill.
• Standardized: Built on the open WebDriver BiDi protocol rather than proprietary alternatives.
• Zero-config: Automatically handles browser downloads and setup.
• Versatile integration: Works as a CLI tool, an MCP server, or via native client libraries for JS/TS, Python, and Java.
• Lightweight: Distributed as a single ~10MB binary with no runtime dependencies.