YetiBrowser MCP

YetiBrowser MCP is a fully open‑source bridge that lets AI assistants—such as Claude Code, Codex, Cursor, or custom tools—to control a real browser tab through the Model Context Protocol (MCP). By coupling a lightweight Node‑based server with Chrome or Firefox extensions, the system keeps all browsing data on the local machine while providing a rich set of automation primitives that are both auditable and easily extensible. This solves the common developer pain point of needing a secure, private browser interface that can be scripted by an AI without exposing user data to external services.

The server exposes a comprehensive suite of browser tools that cover navigation, interaction, and introspection. For example, loads a URL and returns an accessibility‑oriented snapshot; , , and allow fine‑grained DOM manipulation; and highlights changes between page states. Additional utilities such as full‑page screenshots, console log capture, and page‑state dumps give developers visibility into the underlying browser environment—essential for debugging complex web interactions or building QA pipelines. All tools are defined in shared TypeScript schemas, making it straightforward to add new capabilities or adapt existing ones.

Because the extension communicates solely with a localhost MCP server, browsing data never leaves the host device. This local‑first approach is invaluable for privacy‑sensitive applications, internal tooling, or regulated environments where data residency matters. The architecture also supports cross‑browser operation: the same core logic powers both Chrome and Firefox packages (the Firefox build is awaiting full Manifest V3 support), ensuring consistency across platforms.

Integrating YetiBrowser MCP into an AI workflow is simple yet powerful. An assistant can issue a series of tool calls to navigate through a website, fill out forms, or extract information—all while the server maintains sync with the active tab. The rich console capture and DOM diffing features enable developers to verify that each automation step behaved as expected, making it an excellent fit for building reliable web‑automation scripts, automated testing suites, or even conversational agents that need to browse the internet on demand.

Unique advantages of YetiBrowser MCP include its transparency and hackability—there are no opaque binaries, so teams can inspect, fork, or extend every component. The repository layout separates shared definitions, server logic, and extension packaging, which streamlines contributions and custom deployments. Production‑ready tooling is also provided: scripts for packaging extensions, publishing checklists, and integration guides make it straightforward to roll the server out in IDE workflows or continuous‑integration pipelines. Overall, YetiBrowser MCP offers developers a secure, auditable, and feature‑rich bridge between AI assistants and real browser tabs.