Charlotte MCP Server

Browsing starts with a compact map of a page: its landmarks, headings, and counts of interactive elements by region. Agents can then search by text, role, type, or position to obtain element IDs for clicking, entering text, choosing options, or submitting forms. Observation can expand the response to include element metadata, form structure, content summaries, and visible text. CSS selector searches and coordinate clicks provide ways to reach targets outside the accessibility tree, while screenshots offer a visual fallback.

Charlotte also supports tab management and structural comparisons with earlier page snapshots. Saved screenshots can be listed, retrieved, or deleted. Startup profiles select the available tool set, and visibility controls let agents enable or disable groups during a session. Local development tools serve static files with automatic reloads, inject CSS or JavaScript, and run accessibility, performance, SEO, contrast, and broken-link checks. Console output and request logs support inspection alongside those workflows. Page representations cover the main frame; iframe content is not included.