openchrome and chrome-local-mcp
The two tools are competitors, as both aim to provide a Multi-Control Protocol (MCP) server for automating Chrome, but they differ in their approach and intended use-case: `shaun0927/openchrome` focuses on controlling a "real Chrome" instance for AI agents, implying a more general and potentially distributed automation solution, while `callmehuyv/chrome-local-mcp` specifically targets local Chrome automation for "Claude Code" powered by Puppeteer, suggesting a more specialized and constrained local development environment.
About openchrome
shaun0927/openchrome
Open-source browser automation MCP server. Control your real Chrome from any AI agent.
Implements MCP server that connects directly to Chrome via Chrome DevTools Protocol (CDP), eliminating middleware overhead while maintaining real browser sessions and login state across parallel tabs. Includes 27 harness subsystems—hint engine with 30+ rules, outcome classification, pattern learning, and DOM compression achieving 15x token efficiency—that guide AI agents and prevent costly error cascades. Achieves ~300MB memory for 20 parallel sessions versus 5GB+ for traditional approaches, while remaining invisible to bot detection and enabling authenticated cross-site operations.
About chrome-local-mcp
callmehuyv/chrome-local-mcp
A local Chrome browser automation MCP server for Claude Code, powered by Puppeteer
Implements persistent browser sessions with real Chrome profiles, accessibility snapshots (reducing vision token costs), and batch actions—all via a single stdio-based MCP server that requires zero configuration. Includes 22 specialized tools optimized for Claude Code workflows, plus optional local OCR via Apple Vision and a REST API for external integrations.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work