Built-in Tools

MoleClaw includes 37+ built-in tools covering all aspects of browser automation. The AI automatically selects the most suitable tools based on your natural language instructions.

Page Awareness

Tool	Description
page_viewer	Get the current page's visible text content for understanding page information
page_snapshot	Take a structured snapshot of the page, supports query-based element targeting, returns element tree and element_id
page_skeleton	Extract page skeleton structure for a quick overview of the overall layout
page_assert	Assert whether the page state meets expectations, used for verification after critical operations
page_repair	When page_assert fails, attempt to automatically repair the page state
screenshot	Take page screenshots — supports visible area, full page, region, and element screenshots (CDP enhanced). annotate=true mode overlays numbered markers on interactive elements and returns an element_id mapping table for precise targeting
selection_context	Get the text selected by the user on the page along with its surrounding context

Page Operations

Tool	Description
cdp_input	Page interaction operations via CDP (trusted events). Supports click, double click, right click, hover, drag, type, key press, scroll, fill, clear, select, click by text, scroll into view, focus, wait for element/text/navigation, get element info. Two targeting modes: element_id (from page_snapshot, preferred) and CSS selector.
cdp_dom	Cross-origin DOM operations via CDP: query elements, modify attributes, insert/remove nodes, read/write HTML, get precise box model, etc.
cdp_frame	Cross-iframe operations via CDP: list all page frames, execute JS in a specific frame (expression parameter), get iframe text snapshot. Solves cross-origin iframe issues

Navigation & Tabs

Tool	Description
tab_navigate	Tab navigation: navigate URL in current tab, open new tab, close tab, etc.
fetch_url	Make HTTP requests in the background to fetch URL content without affecting the current page

Browser Capabilities

Tool	Description
clipboard_ops	Clipboard operations: read and write clipboard content
storage_kv	Key-value storage: read/write data in the extension's local storage, supports cross-session persistence
notification	Send browser desktop notifications
bookmark_ops	Bookmark operations: search, create, delete bookmarks
history_search	Search browser history
download_file	Download files to local disk

Timers & Automation

Tool	Description
timer	Unified timer management: set delayed tasks, recurring tasks, clear timers, and list active timers
resident_runtime	Resident runner: continuously run tasks in the background, suitable for long-term monitoring or periodic execution

Workflows & Tasks

Tool	Description
site_workflow	Execute predefined site workflows, automatically matches available workflows based on the current page URL
spawn_subtask	Split independent sub-goals into isolated tasks, each with its own context
request_confirmation	Request user confirmation before performing sensitive or irreversible actions. The user sees the message and can approve or reject with an optional reason
ask_user	Ask the user a question to gather missing information or let them choose between options. Supports preset options and/or free text input
save_workflow	Save a user-confirmed workflow definition to the registry, used in the dialog-based recording confirmation flow

CDP Enhanced Tools Chrome DevTools Protocol

Connected to Chrome DevTools Protocol via the chrome.debugger API, providing browser-process-level deep control. These tools are especially critical when regular Content Script approaches are limited (anti-bot detection, cross-origin iframes, network details, etc.).

Input & Interaction

Tool	Description
cdp_input	Send trusted mouse/keyboard events (isTrusted=true): click, double-click, right-click, hover, drag, type text, keypress, scroll. Bypasses anti-bot event source detection
cdp_dialog	Query and handle JavaScript dialogs (alert/confirm/prompt/beforeunload), supports manual handling and auto strategies

Pages & Frames

Tool	Description
cdp_frame	Cross-iframe operations: list all page frames, execute JS in a specific frame, get iframe text snapshot. Solves cross-origin iframe issues (CAPTCHAs, payment forms, etc.)

Network & Cookies

Tool	Description
cdp_network	CDP-enhanced network monitoring: complete request/response data (including body and headers), statistical summaries, and cross-origin Cookie read/write (get/set/delete)

Environment Emulation

Tool	Description
cdp_emulation	Device and environment emulation: viewport size (mobile), User-Agent override, geolocation spoofing, language/timezone settings, network condition simulation (3G/offline, etc.)

Request Interception

Tool	Description
cdp_fetch	Request interception and modification (Fetch domain): intercept page network requests to modify and continue, return custom responses (Mock API), or simulate failures. Useful for auth header injection, data mocking, and CORS bypass

Deep DOM Operations

Tool	Description
cdp_dom	Cross-origin DOM operations (DOM domain): query/modify DOM nodes via CDP, ignoring same-origin policy. Supports CSS selector queries, HTML read/write, attribute operations, precise element box model (margin/border/padding/content), and node removal

TIP

Page storage operations (storage_ prefix actions) and CSS style operations (css_ prefix actions) are now integrated into cdp_dom as unified actions.

Visual Highlighting

Tool	Description
cdp_overlay	Element highlighting (Overlay domain): highlight specified DOM nodes, elements matching CSS selectors, or rectangular regions with custom colors. Visually marks AI operation targets so users can observe what's being operated on

Debug & Diagnostics

Tool	Description
cdp_console	Capture page console.log/warn/error output and uncaught JavaScript exceptions to help AI diagnose page issues

Note

CDP tools require the debugger permission. A debugger notification bar will appear at the top of the browser during use — this is Chrome's security mechanism and is expected behavior.

Tool Selection Priority

When the AI needs to operate on a page, it selects tools in this priority order:

1. site_workflow — Preferred: use predefined workflows when available for the current page, fast and reliable
2. page_snapshot + cdp_input(element_id=...) — Snapshot to locate elements, then operate precisely by element_id
3. cdp_input(selector=...) — CSS selector-based operations when element_id is unavailable
4. cdp_dom — DOM read/write, CSS styles, and page storage operations

TIP

You don't need to manually choose tools — just describe your needs in natural language, and the AI will automatically select the best tool combination for the task.