GitHub ↗

Tools

Browser MCP exposes 40 tools to the connected agent. Every tool acts on the active Chrome tab (or a tab/frame you target explicitly) via the Browser MCP extension — no headless browser, no Playwright binary.

Source of truth: mcp-server/tools.js (TOOLS array). Regenerate this page from that file if tool names, params, or descriptions change.


ToolDescription
browser_navigateNavigate the active tab to a URL; reuses the current tab unless new_tab=true.
browser_get_page_contentReturn the current page's content as text or html.
browser_screenshotScreenshot the visible viewport; returns base64 PNG or saves to a given path.
browser_execute_scriptRun arbitrary JavaScript in the page context and return the result.
browser_extract_listRead every row of a long or virtualised list by scrolling its container until nothing new appears — mail lists, invoice tables, transaction histories. Those UIs keep only a handful of rows in the DOM, so a single page read returns a sliver and looks complete.

Interaction — 14 tools

ToolDescription
browser_clickClick an element via CSS or text selector (text=Submit, button:text(Next)); auto-scrolls into view, uses real mouse events.
browser_double_clickDouble-click an element — for editors and grids that open on double-click rather than single.
browser_right_clickRight-click an element to open its context menu.
browser_click_xyClick at absolute viewport coordinates. Last resort for canvas, maps, and custom-rendered UI where no element can be selected.
browser_fillFill a form input via CSS or text selector; works on CSP-strict sites via the Chrome Debugger API.
browser_press_keySend a keyboard key press (Enter, Tab, Escape, arrows, letters...) with optional ctrl/alt/shift/meta modifiers.
browser_scrollScroll to a matched element, or by a pixel offset.
browser_waitWait for an element matching a CSS or text selector to appear.
browser_hoverHover an element to trigger tooltips, dropdowns, or hover states.
browser_select_optionSelect an option from a native <select> or a custom dropdown (Angular Material, React Select, etc.).
browser_set_comboboxDrive an autocomplete/combobox: click, type filter query, wait for the listbox, click the option(s); supports multi-select chips.
browser_set_dateSet a date input robustly — native value-set, masked-text typing, or calendar-picker navigation (MUI/AntD/react-datepicker/Lexical), with read-back verification.
browser_dismiss_overlaysBulk-dismiss popups, modals, tooltips, and banners via heuristics on close affordances (aria-label, "Skip"/"Ikke nu"/"Got it", × button).
browser_handle_dialogAccept or dismiss a native alert()/confirm()/prompt() dialog; call before the action that triggers it.

Tabs & Frames — 6 tools

ToolDescription
browser_list_tabsList the current session's own open tabs (URL + title) — each concurrent session has its own tab group and can't see another session's tabs.
browser_switch_tabActivate a specific tab by ID.
browser_close_tabClose a tab by ID (session-owned tabs only).
browser_get_new_tabReturn the most recently opened tab — useful after a link opens a new tab or an OAuth popup.
browser_list_framesList all iframes on the current page with URL and index.
browser_select_frameExecute JavaScript inside a specific iframe, targeted by index.

Data & Storage — 5 tools

ToolDescription
browser_get_cookiesGet cookies for a domain.
browser_set_cookiesSet one or more cookies for a domain (single cookie or a cookies[] batch).
browser_get_local_storageRead localStorage from the current page — a single key or all of it.
browser_set_local_storageWrite a localStorage key/value pair on the current page.
browser_console_logsReturn recent console.log/warn/error messages captured from the page.

Files — 2 tools

ToolDescription
browser_upload_fileUpload file(s) to an <input type="file"> via the Chrome Debugger API — no OS file dialog needed.
browser_drop_fileUpload into a drag-drop zone by locating a hidden file input in its subtree or parent (up to 2 levels); use when browser_upload_file finds no input.

Network — 3 tools

ToolDescription
browser_fetchMake an HTTP request from the extension background — not subject to page CORS/CSP.
browser_wait_for_networkWait for a network request matching a URL substring to complete, via Chrome DevTools Protocol.
browser_extract_tokenExtract an API token from any provider's account-settings page; ships zero-config shortcuts (known URL + extraction hint) for 9 common providers (Stripe, HubSpot, Slack, Shopify, Mailchimp, Pipedrive, Calendly, Google, LinkedIn) — any other provider still works via manual navigate + read. Optionally stores the token in the Agent360 vault.

CAPTCHA — 1 tool

ToolDescription
browser_solve_captchaDetect and attempt reCAPTCHA v2/v3, hCaptcha, Cloudflare Turnstile, or FunCaptcha — auto-click checkbox, then AI-vision-guided grid click, then human fallback for the rest.

Human-in-the-Loop — 1 tool

ToolDescription
browser_ask_userShow an overlay asking the user to perform an action or provide input (credentials, 2FA, CAPTCHA, OAuth consent); returns their response.

Meta & Recovery — 3 tools

ToolDescription
browser_provide_feedbackSelf-check plus report in one call. Compares this server against the latest published on npm, the connected extension against this server, and detects more than one Browser MCP extension connected at once — the three things that explain most "it just stopped working" moments. Returns a verdict, concrete fix steps, and a pre-filled issue link for whatever is genuinely missing. The agent calls it on its own whenever a tool blocks it.
browser_aboutReturn Browser MCP info plus pre-filled links for the user to submit a feature wish, share a use-case, or report a bug.
browser_reattach_debuggerForce-detach and re-attach the Chrome debugger on the current tab. Use when click/fill/press_key start timing out or report a ghost attach while browser_list_tabs still works — faster than reloading the extension.

Total: 40 tools (5 + 14 + 6 + 5 + 2 + 3 + 1 + 1 + 3 + 3 = 43), verified against mcp-server/tools.js line-by-line — no invented tools.