IronBee DevTools MCP Server: Setup for Claude, Cursor and Other Apps

• Opens a web address in the browser

  Loads the URL and, by default, returns a text outline of the page with handles you can click or fill.

  Try: “Open https://example.com/login and tell me what is on the page.”

  Network · navigation_go-to

• Goes back or forward in the page history

  Moves one step back or forward in the tab's history and returns the new page outline.

  Try: “Go back to the previous page.”

  Network · navigation_go-back-or-forward

• Reloads the current page

  Refreshes the page and returns the new outline, so you can check a fix after a code change.

  Try: “Reload the page and tell me if the error is gone.”

  Network · navigation_reload

• Outlines the page as a tree of labelled elements

  Returns the page as a text tree of roles and names, with short handles (e1, e2) the click and fill tools accept.

  Try: “Take a snapshot of the checkout form and list its fields.”

  Read · a11y_take-aria-snapshot

• Finds hidden, off-screen or covered elements

  Combines the browser's accessibility tree with size, visibility and overlap checks, to explain why a click misses an element that looks visible.

  Try: “Why can't the Buy button be clicked? Check if something covers it.”

  Read · a11y_take-ax-tree-snapshot

• Returns the page HTML

  Returns the HTML of the whole page or one element. Scripts are stripped unless you ask to keep them.

  Try: “Show me the HTML of the navigation bar.”

  Read · content_get-as-html

• Returns the visible text of the page

  Returns the text a visitor can see, for the whole page or one section, cut off at a length limit.

  Try: “Read me the text of the pricing section.”

  Read · content_get-as-text

• Saves the current page as a PDF file

  Prints the page to a PDF in a folder you choose, or the system temp folder by default.

  Try: “Save this page as a PDF.”

  Write · content_save-as-pdf

• Starts a video recording of the browser

  Begins recording the browser session so you can replay what the assistant did.

  Try: “Start recording, then walk through the signup flow.”

  Write · content_start-recording

• Stops recording and saves the video

  Ends the recording and writes the video file to a folder you choose, then returns its path.

  Try: “Stop the recording and tell me where the video is saved.”

  Write · content_stop-recording

• Takes a screenshot of the page

  Saves a PNG or JPEG of the page or an element to disk and returns the file path. Full-page and annotated shots are options.

  Try: “Take a full-page screenshot of the homepage.”

  Write · content_take-screenshot

• Clicks an element on the page

  Clicks the element you name by handle, text description or CSS selector. It can wait for the next page and return its outline.

  Try: “Click the Sign in button.”

  Write · interaction_click

• Drags one element onto another

  Drags a source element to a target element, for sliders, sortable lists and drag-and-drop uploads.

  Try: “Drag the first card into the Done column.”

  Write · interaction_drag

• Types into a form field

  Fills an input with the text you give it. Combined with click, it can complete and submit forms.

  Try: “Fill the email field with test@example.com.”

  Write · interaction_fill

• Moves the mouse over an element

  Hovers an element, for menus and tooltips that only show on hover.

  Try: “Hover over the Products menu and list what opens.”

  Write · interaction_hover

• Presses a keyboard key

  Sends a key press, such as Enter or Escape, to the page or to a chosen element.

  Try: “Press Escape to close the dialog.”

  Write · interaction_press-key

• Changes the simulated screen size

  Sets the page's viewport width and height, so you can test phone, tablet and desktop layouts.

  Try: “Resize the viewport to 390 by 844 and screenshot the homepage.”

  Write · interaction_resize-viewport

• Resizes the real browser window

  Changes the size of the actual browser window. It works best in Chromium with the simulated viewport turned off.

  Try: “Make the browser window 1280 by 800.”

  Write · interaction_resize-window

• Picks an option in a dropdown

  Chooses a value in a select menu by handle or selector.

  Try: “Choose Canada in the country dropdown.”

  Write · interaction_select

• Scrolls the page or a scrollable box

  Scrolls by an amount, to an element, or to the top, bottom, left or right of the page or a container.

  Try: “Scroll to the bottom of the page.”

  Read · interaction_scroll

• Reads the browser console log

  Returns console messages captured from the page, filtered by level, search text or time.

  Try: “Show me the console errors since the page loaded.”

  Read · o11y_get-console-messages

• Lists the requests the page made

  Returns captured network requests, filterable by type and status. Headers and response bodies are included only if you ask.

  Try: “Which API calls returned a 500 on this page?”

  Read · o11y_get-http-requests

• Measures page speed and layout stability

  Collects Core Web Vitals (LCP, INP, CLS) plus TTFB and FCP, with Google's ratings and suggestions.

  Try: “Check the Core Web Vitals for this page.”

  Read · o11y_get-web-vitals

• Creates a trace ID for this session

  Generates a new OpenTelemetry-style trace ID and attaches it to the session so later requests share it.

  Try: “Start a new trace ID before I test the checkout.”

  Write · o11y_new-trace-id

• Shows the current trace ID

  Returns the trace ID and trace state set for this session, if any.

  Try: “What trace ID is this session using?”

  Read · o11y_get-trace-context

• Sets or clears the trace ID

  Pins a trace ID or trace state you choose on the session, or clears it with an empty value.

  Try: “Set the trace ID to the one from the failing request.”

  Write · o11y_set-trace-context

• Reads one trace back from the IronBee platform

  Fetches every recorded span of one distributed trace as a tree. It reads from the IronBee service, so it needs that service to be set up.

  Try: “Show me the full trace for the last request.”

  Network · o11y_get-trace

• Reads all traces of this session

  Fetches all spans recorded under the current IronBee session, optionally narrowed to one activity.

  Try: “List every trace recorded in this session.”

  Network · o11y_get-session-traces

• Reads the logs linked to one trace

  Fetches the browser console output and app log lines recorded under one trace ID from the IronBee platform.

  Try: “Show the logs that belong to this trace.”

  Network · o11y_get-trace-logs

• Reads all logs of this session

  Fetches every log line recorded under the current IronBee session, optionally narrowed to one activity.

  Try: “Show all log lines from this session.”

  Network · o11y_get-session-logs

• Finds the React component behind an element

  Looks up which React component renders a DOM element, using React internals. Results are best-effort and vary by build.

  Try: “Which React component renders the header banner?”

  Read · react_get-component-for-element

• Finds the page elements a React component draws

  Maps a React component to the DOM elements it renders, using React internals. Works best with the React DevTools extension.

  Try: “Where on the page does the CartSummary component show up?”

  Read · react_get-element-for-component

• Maps minified code to original source

  Uses source maps to turn a bundle URL, line and column into the original file, line and column.

  Try: “Where in the source is line 1 column 4521 of bundle.min.js?”

  Read · debug_resolve-source-location

• Sets a probe that records state without pausing

  Adds a tracepoint at a script location that captures the call stack and local variables each time it runs, without stopping the page.

  Try: “Put a tracepoint at line 42 of checkout.js.”

  Write · debug_put-tracepoint

• Sets a probe that logs a value

  Adds a logpoint at a script location that evaluates an expression you give it each time that line runs.

  Try: “Log the cart total whenever line 88 of cart.js runs.”

  Write · debug_put-logpoint

• Chooses when errors are captured

  Sets whether to capture snapshots on no exceptions, uncaught ones only, or all of them.

  Try: “Capture a snapshot on every uncaught exception.”

  Write · debug_put-exceptionpoint

• Adds an expression to evaluate at each hit

  Adds a watch expression that is evaluated on every tracepoint hit and shown in the snapshot.

  Try: “Watch user.name at every tracepoint.”

  Write · debug_add-watch

• Lists tracepoints, logpoints and watches

  Returns the probes and watch expressions currently set, optionally only one kind.

  Try: “List my active probes.”

  Read · debug_list-probes

• Removes one probe or watch

  Removes a tracepoint, logpoint or watch expression by its ID.

  Try: “Remove the tracepoint on checkout.js.”

  Delete · debug_remove-probe

• Removes all probes of a kind

  Removes all tracepoints, logpoints and watches, or only the kinds you name.

  Try: “Clear all probes.”

  Delete · debug_clear-probes

• Reads what the probes captured

  Returns captured snapshots from tracepoints, logpoints and exception captures, trimmed to the top call frame and local variables by default.

  Try: “What did the tracepoint capture?”

  Read · debug_get-probe-snapshots

• Deletes captured probe snapshots

  Clears stored snapshots for all probe types or the ones you name, or for one probe.

  Try: “Clear the captured snapshots.”

  Delete · debug_clear-probe-snapshots

• Saves a reusable script

  Stores a JavaScript scenario that can call other tools, on disk in the project folder or globally.

  Try: “Save this login flow as a scenario called login.”

  Write · scenario-add

• Edits a saved script

  Changes a saved scenario's description, script, parameters or metadata. A metadata merge can remove keys.

  Try: “Update the login scenario to use the new button label.”

  Delete · scenario-update

• Deletes a saved script

  Deletes a scenario by name from the project or global folder.

  Try: “Delete the old-login scenario.”

  Delete · scenario-delete

• Lists saved scripts

  Lists the scenarios in the project and global folders, optionally filtered by their metadata.

  Try: “What scenarios do I have saved?”

  Files · scenario-list

• Searches saved scripts by description

  Searches scenario names and descriptions across project and global scenarios and ranks the matches.

  Try: “Find a scenario that logs in.”

  Files · scenario-search

• Lists tool calls recorded during a scenario step

  While a scenario run is paused for the assistant, returns the browser tool calls made since the pause began.

  Try: “Show what you did since the scenario paused.”

  Read · scenario-recorded-actions

• Runs a saved script

  Runs a saved scenario by name. It uses the same JavaScript sandbox as the execute tool and can call any other tool.

  Try: “Run the login scenario.”

  Runs code · scenario-run

• Changes requests before they are sent

  Adds a rule that adds or overrides headers or method on requests matching a URL pattern.

  Try: “Add a test header to all calls to /api/*.”

  Write · stub_intercept-http-request

• Fakes the reply to a request

  Adds a rule that answers matching requests with a canned status, headers and body, or blocks them, with optional delay.

  Try: “Make /api/user return a 500 error.”

  Write · stub_mock-http-response

• Lists active request rules

  Returns the intercept and mock rules currently installed in the session.

  Try: “Which request rules are active?”

  Read · stub_list

• Removes request rules

  Removes one rule by ID, or all rules for the session when no ID is given.

  Try: “Remove all mock rules.”

  Delete · stub_clear

• Waits until the page stops loading

  Waits until the page has few requests in flight for a set time, so screenshots and snapshots are stable.

  Try: “Wait for the page to finish loading, then screenshot it.”

  Read · sync_wait-for-network-idle

• Runs JavaScript that can call other tools

  Runs a script you or the assistant write in a restricted sandbox with the Playwright page object. It can call any other tool up to 50 times and run code in the page.

  Try: “Run a script that logs in, opens the settings page and returns its title.”

  Runs code · execute

• Used by the assistant behind the scenes

  Shows the state of debugging

  Reports whether debugging is on and how many tracepoints, logpoints, watches and snapshots exist.

  Try: “What is the debugging status?”

  Read · debug_status

• Waits for background reporting to finish

  Waits for queued background work such as event and artifact uploads to settle before a run ends.

  Try: “Flush pending uploads before we finish.”

  Read · flush

• Lists the secrets given to this run

  Lists secret names and the placeholder to use in tool input. It never returns the secret values.

  Try: “Which secrets can I use in this session?”

  Read · secrets_list

Watch out for

• It runs scripts with full control of the page

  The execute tool and saved scenarios run JavaScript with the Playwright page, so they can click, submit forms and read anything in a logged-in session. There is no read-only mode, so only open sites you trust. npmjs.com

• It can click, type and submit in logged-in sessions

  Click, fill and select act like you. Use a separate browser profile for tests and keep real accounts out of it. npmjs.com

• It can delete saved scripts and probes

  scenario-delete removes a saved script file, and scenario-update can drop metadata keys. The probe and request-rule clear tools only remove in-memory debug state. npmjs.com

• It writes screenshots, PDFs and videos to disk

  Screenshots, PDFs and recordings go to the folder you give or the system temp folder. Scenario scripts are stored as files under the project folder or a global folder. npmjs.com

• It sends anonymous usage events by default

  The package sends tool names, durations, success flags, client name, OS and Node version to PostHog unless you set TELEMETRY_ENABLE=false. It also downloads browsers on first run. npmjs.com

• Optional Jira and Linear tools are off by default

  They appear only when you set Jira or Linear credentials, and the write tools need a separate write flag. npmjs.com

Security advisories