Element To Llm MCP server

What it does

E2LLM is a perception layer for AI in web browsers that captures live pages as structured, LLM-readable documents (SiFR format) and performs explicit, gated browser actions. It works as a remote MCP server via a browser extension, providing AI with machine-readable understanding of web pages and the ability to interact with them—without autonomous decision-making.

• Capture and analyze the current state of a web page in structured format to understand what is displayed and what actions are available.
• Fill forms, click buttons, type text, or perform other explicit interactions on web pages based on AI reasoning, with each action individually approved.
• Query and filter page elements by tag, salience, text, or selector to locate specific content without re-capturing the entire page.

Tools this server exposes

Tool	Does
sifr_capture	Capture the current page as a SiFR document—a compact, structured, LLM-readable model of page content and available actions.
query	Filter the current capture by tag, salience, text, or selector to locate specific elements.
inspect	Get full detail for a single element including selector, attributes, computed styles, and box geometry.
read_page	Read the page text as markdown in reading order.
list_tabs	List all open browser tabs.
act	Perform one explicit interaction: click, type, select, navigate, drag, open a new tab, or paste.
batch_act	Run a planned sequence of interactions (click, type, select, drag, paste) then observe the result once.
explore	Scroll, hover, or recapture to read more of the page without changing its state.
close_tab	Close a browser tab.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.