browser_evaluate
Execute arbitrary JavaScript code in the page context. Has full access to the DOM, window object, and all page JavaScript. Use return_value=true to get the result of an expression.
When to use browser_evaluate
Use browser_evaluate when you need to run custom JavaScript inside the page context. It is part of Owl Browser's JavaScript Evaluation toolset and runs inside a self-hosted, source-level stealth engine, so every call inherits the same undetectable browser fingerprint as the rest of your automation — no separate anti-detect setup required.
Usage Example
Parameters
Required
context_idstringrequiredThe unique identifier of the browser context (e.g., 'ctx_000001')
Optional
scriptstringJavaScript code to execute in the page context. Can access DOM, window object, etc. Example: 'document.title' or 'window.scrollY'. Optional if expression is provided.
expressionstringJavaScript expression to evaluate and return (shorthand for script with return_value=true). If provided, script is optional and return_value is automatically true.
return_valuebooleanIf true, treats the script as an expression and returns its value. If false (default), executes the script as statements. Automatically set to true when using expression.
Response
Returns a JSON object with the operation result.
{
"success": true,
"result": <value>
}Frequently Asked Questions
What does browser_evaluate do?
Execute arbitrary JavaScript code in the page context. Has full access to the DOM, window object, and all page JavaScript. Use return_value=true to get the result of an expression. It belongs to Owl Browser's JavaScript Evaluation category and is available through the REST API, the Python SDK (browser.evaluate()), the Node.js SDK, and the MCP server.
What parameters does browser_evaluate accept?
browser_evaluate accepts 1 required parameter (context_id) and 3 optional parameters. All parameters are sent as JSON in a POST request to /api/execute/browser_evaluate.
Is browser_evaluate detectable by anti-bot systems like Cloudflare or DataDome?
No. browser_evaluate executes inside Owl Browser's Chromium engine, which applies fingerprint spoofing at the C++ source level rather than through JavaScript patches. Every tool call shares the same consistent, human-like fingerprint, so anti-bot systems such as Cloudflare, DataDome, and Akamai see an ordinary browser.
Related Tools
browser_create_contextCreate a new isolated browser context with its own cookies, storage, and optional proxy configuration. Each context acts as an independent browser session. Use this to create multiple isolated browsing sessions, configure proxy/Tor connections, load browser profiles with saved fingerprints, and enable/disable LLM features. Returns a context_id to use with other browser tools.
browser_navigateNavigate the browser to a specified URL. This is a non-blocking operation that starts navigation and returns immediately. Use browser_wait_for_network_idle or browser_wait_for_selector to wait for the page to fully load. Supports HTTP, HTTPS, file, and data URLs. When wait_until is set (load, networkidle, fullscroll, domcontentloaded) and the page declares WebMCP tools, the response includes a webmcp_tools array containing the full tool definitions (name, description, inputSchema). Use browser_webmcp_call_tool to execute any of these tools directly.
browser_observeAgent-native page observation. Returns the compacted OwlMark render (text-only structural view of the page), a handle table of interactive elements with stable tokens, page metadata, and a token estimate. Pass a handle token (e.g. 'b3') or 'pm:N' to browser_click/browser_type. Requires the context to be created with render_mode 'agent' or 'both'. ~20-100x fewer tokens than a screenshot for AI agent page understanding.
browser_clickClick on an element using CSS selector, XY coordinates, or natural language description. Supports semantic element finding using AI - describe what you want to click (e.g., 'login button', 'search icon') and the system will locate the right element. Simulates a real mouse click with proper event dispatch. Optionally hold the mouse button for press-and-hold interactions using hold_ms.
browser_typeType text into an input field with human-like keystroke simulation. Target the field using CSS selector, coordinates, or natural language (e.g., 'email field'). When selector is omitted, types into the currently focused element. Note: Does NOT clear existing content - use browser_clear_input first if you need to replace text rather than append.
browser_get_page_mapGet a compact structured map of all interactive elements on the page, grouped by sections. Returns a markdown table with numbered references, types, descriptions, current values, and selectors. Use element numbers as selectors in browser_click and browser_type (e.g., selector '5'). 10x cheaper than screenshots for AI agent page understanding.