{ "tools": [ { "name": "navigate", "description": "Navigate to URLs or move through browser history.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "The browser tab to navigate in." }, "url": { "type": "string", "description": "The URL to navigate to, or \"back\"/\"forward\" for history navigation." } }, "required": [ "tab_id", "url" ] }, "usage": [ "navigate(url=\"https://example.com\", tab_id=123)", "navigate(url=\"back\", tab_id=123)", "navigate(url=\"forward\", tab_id=123)" ], "best_practices": [ "Always include the tab_id parameter.", "URLs can be provided with or without protocol; default to https:// when omitted.", "Use for loading new web pages or navigating between pages." ] }, { "name": "computer", "description": "Interact with the browser through mouse clicks, keyboard input, scrolling, and screenshots.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "Browser tab to interact with." }, "action": { "type": "string", "description": "Action to perform.", "enum": [ "left_click", "right_click", "double_click", "triple_click", "type", "key", "scroll", "screenshot" ] }, "coordinate": { "type": "array", "description": "[x, y] coordinates for mouse or scroll actions.", "items": { "type": "number" }, "minItems": 2, "maxItems": 2 }, "text": { "type": "string", "description": "Text to type or keyboard shortcut to press." }, "scroll_parameters": { "type": "object", "description": "Parameters for scroll actions, such as direction and amount." } }, "required": [ "tab_id", "action" ] }, "action_types": [ "left_click", "right_click", "double_click", "triple_click", "type", "key", "scroll", "screenshot" ], "examples": [ "left_click: coordinate=[x, y]", "type: text=\"Hello World\"", "key: text=\"ctrl+a\" or text=\"Return\"", "scroll: coordinate=[x, y], scroll_parameters={\"scroll_direction\": \"down\", \"scroll_amount\": 3}" ] }, { "name": "read_page", "description": "Extract page structure and get element references from the DOM accessibility tree.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "Browser tab to read." }, "depth": { "type": "integer", "description": "How deep to traverse the tree.", "default": 15 }, "filter": { "type": "string", "description": "Element filter mode.", "enum": [ "interactive", "all" ] }, "ref_id": { "type": "string", "description": "Focus on a specific element's children." } }, "required": [ "tab_id" ] }, "returns": [ "Element references such as ref_1 and ref_2.", "Element properties, text content, and hierarchy." ], "best_practices": [ "Use when screenshot-based clicking might be imprecise.", "Get element references before using form_input or computer tools.", "Use smaller depth values if output is too large.", "Filter for interactive when only interested in clickable elements." ] }, { "name": "find", "description": "Search for elements using natural language descriptions.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "Browser tab to search in." }, "query": { "type": "string", "description": "Natural language description of what to find, such as search bar or add to cart button." } }, "required": [ "tab_id", "query" ] }, "returns": [ "Up to 20 matching elements with references and coordinates." ], "best_practices": [ "Use when elements are not visible in the current screenshot.", "Provide specific, descriptive queries.", "Use after read_page if that tool's output is incomplete.", "Use returned references or coordinates with other tools." ] }, { "name": "form_input", "description": "Set values in form elements, including text inputs, dropdowns, and checkboxes.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "Browser tab containing the form." }, "ref": { "type": "string", "description": "Element reference from read_page, such as ref_1." }, "value": { "description": "Value to set; string for text/dropdowns or boolean for checkboxes." } }, "required": [ "tab_id", "ref", "value" ] }, "usage": [ "form_input(ref=\"ref_5\", value=\"example text\", tab_id=123)", "form_input(ref=\"ref_8\", value=true, tab_id=123)", "form_input(ref=\"ref_12\", value=\"Option Text\", tab_id=123)" ], "best_practices": [ "Always get element refs from read_page first.", "Use for accurate form completion.", "Can handle multiple field updates in sequence." ] }, { "name": "get_page_text", "description": "Extract raw text content from the page.", "parameters": { "type": "object", "properties": { "tab_id": { "type": "integer", "description": "Browser tab to extract text from." } }, "required": [ "tab_id" ] }, "returns": [ "Plain text content without HTML formatting.", "Article or main content when available." ], "best_practices": [ "Use for long articles or text-heavy pages.", "Combine with other tools for comprehensive page analysis.", "For infinite scroll pages, scroll to load all content before extracting." ] }, { "name": "search_web", "description": "Search the web for current and factual information.", "parameters": { "type": "object", "properties": { "queries": { "type": "array", "description": "Keyword-based search queries, maximum 3 per call.", "items": { "type": "string" }, "maxItems": 3 } }, "required": [ "queries" ] }, "returns": [ "Search results with titles, URLs, snippets, and citation IDs." ], "best_practices": [ "Use short, keyword-focused queries.", "Use at most 3 queries per call.", "Break multi-entity questions into separate queries.", "Use this instead of navigating to Google.com.", "Prefer queries such as 'inflation rate Canada' over full questions." ] }, { "name": "tabs_create", "description": "Create new browser tabs.", "parameters": { "type": "object", "properties": { "url": { "type": "string", "description": "Starting URL for the new tab.", "default": "about:blank" } } }, "returns": [ "New tab ID for use with other tools." ], "best_practices": [ "Use for parallel work on multiple tasks.", "Each tab maintains its own state.", "Check tab context after creation." ] }, { "name": "todo_write", "description": "Create and manage task lists.", "parameters": { "type": "object", "properties": { "todos": { "type": "array", "description": "Todo items.", "items": { "type": "object", "properties": { "content": { "type": "string", "description": "Imperative form, such as Run tests." }, "status": { "type": "string", "enum": [ "pending", "in_progress", "completed" ] }, "active_form": { "type": "string", "description": "Present continuous form, such as Running tests." } }, "required": [ "content", "status", "active_form" ] } } }, "required": [ "todos" ] }, "best_practices": [ "Use for tracking progress on complex tasks.", "Mark tasks as completed immediately when done.", "Update frequently to show progress." ] } ], "tool_calling_best_practices": { "proper_parameter_usage": [ "Always include tab_id when required by the tool.", "Provide parameters in the correct order.", "Use JSON format for complex parameters.", "Double-check parameter names match tool specifications." ], "efficiency_strategies": [ "Combine multiple actions in a single computer call when possible.", "Use read_page before clicking for more precise targeting.", "Avoid repeated screenshots when another tool provides the same data.", "Use find when elements are not in the latest screenshot.", "Batch form inputs when completing multiple fields." ], "error_recovery": [ "Take a screenshot after a failed action.", "Re-fetch element references if the page changed.", "Verify the tab_id still exists.", "Adjust coordinates if elements moved.", "Use a different tool approach if the first attempt fails." ], "coordination_between_tools": [ "read_page -> get element refs.", "computer -> interact with elements.", "form_input -> set form values with refs.", "get_page_text -> extract content after navigation.", "navigate -> load new pages before other interactions." ] }, "common_tool_sequences": { "navigating_and_reading": [ "navigate to URL", "wait for page load", "screenshot to see current state", "get_page_text or read_page to extract content" ], "form_completion": [ "navigate to form page", "read_page to get form field references", "form_input for each field", "find or read_page to locate submit button", "computer left_click to submit" ], "web_search": [ "search_web with relevant queries", "navigate to promising results", "get_page_text or read_page to verify information", "extract and synthesize findings" ], "element_clicking": [ "screenshot to see page", "use coordinates with computer left_click or read_page references with computer left_click" ] } }