eze-is/web-access

What it solves

Web-Access provides AI agents with comprehensive web connectivity and browser automation capabilities. It addresses the limitations of basic web search and fetch tools by adding scheduling strategies, direct browser control via the Chrome DevTools Protocol (CDP), and the ability to accumulate site-specific operational experience to improve future interactions.

How it works

The project implements a CDP Proxy that connects directly to a user's existing browser (Chrome or Edge) via WebSocket. This allows the agent to operate within the user's active session, inheriting login states and handling dynamic content. It provides a set of HTTP APIs for the agent to perform actions such as creating new tabs, executing JavaScript, simulating real mouse clicks, uploading files, and taking screenshots. The system also includes a mechanism to search local browser bookmarks and history to find internal or previously visited URLs.

Who it’s for

Developers and users of AI agents that support the SKILL.md format (such as Claude Code, Cursor, Gemini CLI, and Codex CLI) who want their agents to have full, human-like web browsing and interaction capabilities.

Highlights

• CDP Proxy Integration: Connects to Chrome or Edge to use existing login sessions and handle dynamic pages.
• Automatic Tool Selection: Intelligently chooses between WebSearch, WebFetch, curl, Jina, and CDP based on the scenario.
• Local Resource Retrieval: Can search through local browser bookmarks and history to locate URLs not found on the public web.
• Site Experience Accumulation: Stores and reuses operational experience (URL patterns, platform characteristics) across sessions by domain.
• Parallel Execution: Supports dispatching sub-agents to perform parallel research across multiple targets with tab-level isolation.
• Advanced Interaction: Supports multiple click methods (JS click and real mouse events) and file uploads.