Navigate to a URL. Must be called before other browser tools.

Returns a compact page snapshot with interactive elements and ref IDs (like @e1, @e2) — no need to call browser_snapshot separately.

Args: url: URL to navigate to session_id: Browser session ID (default: "default")

Get a text snapshot of the current page's accessibility tree.

Returns interactive elements with ref IDs (like @e1, @e2) for use with browser_click and browser_type.

Args: session_id: Browser session ID full: If true, return full page content. If false (default), return only interactive elements (compact view).

Click an element by its ref ID from the snapshot.

Args: ref: Element reference like "@e5" or "@e12" from the snapshot session_id: Browser session ID

Type text into an input field identified by its ref ID.

Clears the field first, then types the new text.

Args: ref: Element reference like "@e3" from the snapshot text: Text to type into the field session_id: Browser session ID

Press a keyboard key.

Args: key: Key to press (e.g., "Enter", "Tab", "Escape", "ArrowDown", "Control+a", "Meta+a") session_id: Browser session ID

Scroll the page up or down.

Args: direction: "up" or "down" pixels: Number of pixels to scroll (default: 500) session_id: Browser session ID

Go back in browser history.

Args: session_id: Browser session ID

列出当前 session 缓冲区里已留痕的页面操作步骤。

每次调 browser_navigate/click/type/press/scroll/back 都会自动留痕。 用这个 tool 看看刚才干了哪些步骤，再决定导出哪些（配合 browser_export_steps 的 from_step 参数）。

Returns: 步骤列表，每行一个步骤：编号 + action + role+name + value

把缓冲区的页面操作步骤导出成 jsonl 测试用例，同时保存登录态。

导出后清空缓冲区。以后回归时：让 agent Read 这个 jsonl， 用 browser state load 加载 auth.json，按步骤用 browser_navigate/click/type 重放 （用 role+name 定位，不用 ref），看当下 url/title 跟录制时一不一致自己判断对错。

Args: name: 用例名，文件名为 test_{name}.jsonl domain: 领域名（如 auth/task/file/cron/skill），作为子目录名 from_step: 从第几步开始导出（1-based），默认 1=全部。用 browser_list_steps 看编号后决定。 session_id: Browser session ID（用于保存登录态）

Take a screenshot of the current page.

Use this when you need to see what the page looks like visually — CAPTCHAs, complex layouts, or when the text snapshot is not enough.

Args: session_id: Browser session ID full_page: If true, capture the full page (not just viewport)

Get browser console output or evaluate a JavaScript expression.

Without expression: returns recent console.log/warn/error messages. With expression: evaluates the JS and returns the result.

Args: session_id: Browser session ID expression: JavaScript expression to evaluate (optional)

List all images on the current page with their URLs and alt text.

Args: session_id: Browser session ID

Get network request logs from the current browser session.

Returns captured HTTP requests with URL, method, status code, headers, resource type, and timing. Useful for discovering hidden APIs, debugging failed requests, and analyzing page load performance.

Args: session_id: Browser session ID url_filter: Filter by URL pattern (e.g., "api", "json", ".js") method: Filter by HTTP method (e.g., "GET", "POST") status: Filter by status code (e.g., "200", "4xx", "404") resource_type: Filter by resource type (e.g., "xhr", "fetch", "document")

Get the full details of a specific network request, including response body.

Use browser_get_network_logs first to find request IDs, then call this tool to see the complete request/response including body content.

Args: request_id: The request ID from browser_get_network_logs (e.g., "39C8D9...") session_id: Browser session ID

Intercept, block, or mock network requests matching a URL pattern.

Use this to:

• Block requests (e.g., block ads, tracking scripts)

• Mock API responses (return custom JSON)

• Test error handling (simulate network failures)

Args: url_pattern: URL pattern to match (e.g., "/api/*", "/analytics/*") action: "block" to abort matching requests, "mock" to return custom response response_body: JSON string for mock responses (only used when action="mock") session_id: Browser session ID