5 items
Turn your Chrome into an MCP server — AI sees the page as numbered markdown and controls it by element ID. Oya Browser turns your Chrome into an MCP server. AI tools like Cursor, Claude Code, and Claude Desktop can see your page as numbered markdown and control it by element ID. No headless browsers. No puppeteer. No screenshots. Your real Chrome — your cookies, your logins, your extensions. The AI reads clean markdown and says click(5) or type(9, "hello"). That's it. HOW IT WORKS 1. Install this extension 2. Run the local bridge: npx oya-browser-mcp 3. Add the MCP config to your AI tool 4. Done — AI can now control your browser The extension analyzes the page, numbers every interactive element, and returns structured markdown. Your AI tool connects via MCP and controls the browser by element number. Everything runs locally on your machine. MCP TOOLS Page: • analyze_page — full page as markdown with numbered interactive elements • screenshot — capture visible tab as PNG • scroll — scroll up or down • wait — wait for a CSS selector to appear • navigate — go to any URL Interaction: • click — click element by number • type — type text into input by number • press_key — press Enter, Escape, Tab, arrows, etc. • click_coordinates — click at x,y pixels • double_click — double-click element or coordinates • mouse_move — hover to trigger tooltips and dropdowns • keyboard_type — type into whatever is focused • drag — drag from one point to another Tabs: • list_tabs — list all open tabs • open_tab — open a new tab • switch_tab — switch to a tab by ID • close_tab — close a tab WHAT THE AI SEES --- url: https://github.com/settings/profile title: Your Profile viewport: 1440x900 elements: 34 total, 18 visible --- # Public profile Name: [#1 input:text value="mk"] Bio: [#2 textarea placeholder="Tell us about yourself"] [#4 button "Update profile"] type(element_id=1, text="new name") then click(element_id=4). No CSS selectors. No HTML parsing. Just the number. WORKS WITH • Cursor — paste MCP config into Settings → MCP • Claude Code — add to your MCP config • Claude Desktop — Settings → Developer → Edit Config • Any MCP client that supports Streamable HTTP SETUP MCP config to paste into your AI tool: { "mcpServers": { "oya-browser": { "url": "http://localhost:9333/mcp" } } } Start the bridge: npx oya-browser-mcp WHY NOT PUPPETEER / PLAYWRIGHT / BROWSER-USE? • Real browser — your Chrome, your cookies, your logins. Not a new browser that starts fresh. • Invisible — it IS a real browser. No bot detection. • Simple — click(5) not page.click('div.css-1a2b > span:nth-child(3)') • Lightweight — 32KB extension + tiny bridge. Not a 400MB Chromium binary. • Standard — MCP protocol. Works with any AI tool, not locked to one SDK. SIDE PANEL Click the extension icon to toggle the built-in side panel: • Run analyze, screenshot, scroll, click, type directly • View the page source as the AI sees it • Test element interactions before your agent runs them PRIVACY & SECURITY • Everything runs locally — no data leaves your machine • No accounts, no API keys, no remote servers • The bridge runs on localhost only • No data is collected, stored, or sent anywhere • Open source: https://github.com/OyadotAI/AgentChrome SUPPORT Questions or issues: mk@oya.ai Questions or issues: mk@oya.ai GitHub: https://github.com/OyadotAI/AgentChrome Built by Oya.ai
Apr 8, 2026
rating_count is the Chrome Web Store ratings count, not a written-review count.
Media assets
Screenshots and videos on the listing.
Has promo video
Whether the listing includes at least one video.
Languages
Declared language locales.
Developer website
Listing exposes a developer website URL.
Contact email
Listing exposes a contact email.
Keyword in name
Case-insensitive substring match in the name.
Keyword in description
Case-insensitive substring match in the description.
Keyword occurrences in description
Count of case-insensitive occurrences in the description.
Category user-count percentile
Share of same-category extensions with fewer users (null if unknown).
These are transparent listing completeness / keyword signals, not a prediction of Chrome Web Store search ranking.