Browser Tools

All browser tools execute locally on your machine via Playwright. They operate on a shared browser instance that persists across calls within a session.

Each browser tool uses an action parameter to group related operations. Most actions automatically return a page snapshot so your AI assistant always has current context.

browser_navigate

Navigate the browser: go to a URL, go back, or go forward.

Parameter	Type	Required	Description
action	enum	No	navigate (default), go_back, or go_forward
url	string	Conditional	URL to navigate to (required for navigate)
return_snapshot	boolean	No	Return page snapshot (default: true)
include_debug	boolean	No	Include console errors and failed network requests (default: false)

// Navigate to a URL
{ "url": "https://myapp.example.com/login" }

// Go back
{ "action": "go_back" }

// Go forward
{ "action": "go_forward" }

browser_interact

Interact with page elements: click, hover, type (keystroke-by-keystroke), fill (replace value), select a dropdown option, or press a key.

Parameter	Type	Required	Description
action	enum	Yes	click, hover, type, fill, select_option, or press_key
selector	string	Conditional	CSS selector (required for click, hover, type, fill, select_option)
value	string	Conditional	Value to type/fill, option to select, or key to press (required for type, fill, select_option, press_key)
slowly	boolean	No	Type with 100ms delay between keystrokes (only for type, useful for autocomplete)
submit	boolean	No	Press Enter after typing (only for type)
return_snapshot	boolean	No	Return page snapshot (default: true)
include_debug	boolean	No	Include console errors and failed network requests

// Click an element
{ "action": "click", "selector": "#login-button" }

// Hover over a dropdown trigger
{ "action": "hover", "selector": ".dropdown-trigger" }

// Type keystroke-by-keystroke (good for autocomplete)
{ "action": "type", "selector": "#search", "value": "test", "slowly": true }

// Fill a field (replaces existing value)
{ "action": "fill", "selector": "#email", "value": "[email protected]" }

// Select a dropdown option
{ "action": "select_option", "selector": "#country", "value": "US" }

// Press a key
{ "action": "press_key", "value": "Enter" }

Note

type vs fill: Use type when you need keystroke-by-keystroke input (autocomplete, search-as-you-type). Use fill to replace the entire field value at once (faster for regular form fields).

browser_form

Fill multiple form fields at once or upload files.

Parameter	Type	Required	Description
action	enum	Yes	fill_form or file_upload
fields	object	Conditional	Map of CSS selector to value (required for fill_form)
selector	string	Conditional	CSS selector of the file input (required for file_upload)
paths	string[]	Conditional	Absolute file paths to upload (required for file_upload)
return_snapshot	boolean	No	Return page snapshot (default: true)
include_debug	boolean	No	Include console errors and failed network requests

// Fill multiple fields at once
{
  "action": "fill_form",
  "fields": {
    "#email": "[email protected]",
    "#password": "secret123",
    "#name": "Test User"
  }
}

// Upload a file
{
  "action": "file_upload",
  "selector": "input[type=file]",
  "paths": ["/Users/me/photo.jpg"]
}

Tip

fill_form is more efficient than multiple browser_interact fill calls — it handles all fields in a single round-trip.

browser_observe

Observe the current page state: get an accessibility snapshot, take a screenshot, view console logs, or inspect network requests.

Parameter	Type	Required	Description
action	enum	Yes	snapshot, screenshot, console_logs, or network_requests
full_page	boolean	No	Capture full scrollable page (only for screenshot, default: false)
filter_status	number	No	Only show requests with this HTTP status or higher (only for network_requests, e.g. 400)

// Get accessibility tree
{ "action": "snapshot" }

// Take a screenshot
{ "action": "screenshot", "full_page": true }

// Get console logs (last 100 messages)
{ "action": "console_logs" }

// Get network requests (errors only)
{ "action": "network_requests", "filter_status": 400 }

Note

Screenshots are returned as MCP image content (image/jpeg). Your AI assistant can see and analyze them directly. The snapshot action returns the page’s accessibility tree — useful for understanding page structure without visual rendering.

browser_assert

Run assertions or wait for elements on the live page.

Parameter	Type	Required	Description
action	enum	Yes	assert or wait
type	enum	Conditional	Assertion type (required for assert — see below)
selector	string	Conditional	CSS selector (for element assertions and wait)
text	string	Conditional	Expected text
url	string	Conditional	Expected URL
count	number	Conditional	Expected element count
attribute	string	Conditional	Attribute name
value	string	Conditional	Expected attribute value
expression	string	Conditional	JavaScript expression (for evaluate_truthy)
timeout_ms	number	No	Timeout in milliseconds (default: 10000, for wait)
return_snapshot	boolean	No	Return page snapshot (default: true)
include_debug	boolean	No	Include console errors and failed network requests

Assertion types:

Type	Required Params	Description
element_visible	selector	Assert element is visible
element_hidden	selector	Assert element is hidden
text_contains	selector, text	Assert element text contains value
text_equals	selector, text	Assert element text equals value exactly
url_contains	text	Assert current URL contains string
url_equals	url	Assert current URL equals value exactly
element_count	selector, count	Assert number of matching elements
attribute_value	selector, attribute, value	Assert element attribute equals value
evaluate_truthy	expression	Assert JavaScript expression evaluates truthy

// Assert login page URL
{ "action": "assert", "type": "url_contains", "text": "/login" }

// Assert button is visible
{ "action": "assert", "type": "element_visible", "selector": "#submit-btn" }

// Assert heading text
{ "action": "assert", "type": "text_equals", "selector": "h1", "text": "Welcome Back" }

// Assert JS expression
{ "action": "assert", "type": "evaluate_truthy", "expression": "document.querySelectorAll('.item').length > 0" }

// Wait for element
{ "action": "wait", "selector": ".loading-spinner", "timeout_ms": 15000 }

// Wait 2 seconds
{ "action": "wait", "timeout_ms": 2000 }

browser_session

Manage browser sessions: save cookies and localStorage, restore a saved session, or open a headed browser for interactive OAuth/SSO login.

Parameter	Type	Required	Description
action	enum	Yes	save, restore, or login
name	string	Yes	Session name (e.g., admin, user, default)
url	string	Conditional	URL to navigate to (required for login)
return_snapshot	boolean	No	Return page snapshot after restore/login (default: false)

// Save current session
{ "action": "save", "name": "admin" }

// Restore a saved session
{ "action": "restore", "name": "admin" }

// Interactive login (opens headed browser)
{ "action": "login", "name": "admin", "url": "https://myapp.example.com/login" }

Sessions are stored locally at ~/.fasttest/sessions/{orgSlug}/{name}.json with restrictive permissions.

Tip

The login action opens a visible browser window so you can complete OAuth/SSO flows manually. Once done, the session is saved and loaded automatically — future runs restore it with the restore action.

browser_page

Page-level operations: manage tabs, resize the viewport, execute JavaScript, handle dialogs, or drag-and-drop.

Parameter	Type	Required	Description
action	enum	Yes	tabs, resize, evaluate, handle_dialog, or drag
tab_action	enum	Conditional	list, create, switch, or close (for tabs)
url	string	No	URL for new tab (for tabs/create)
index	number	No	Tab index (for tabs/switch and tabs/close)
width	number	Conditional	Viewport width in pixels (for resize)
height	number	Conditional	Viewport height in pixels (for resize)
expression	string	Conditional	JavaScript expression to evaluate (for evaluate)
dialog_action	enum	Conditional	accept or dismiss (for handle_dialog)
prompt_text	string	No	Text for prompt dialogs (for handle_dialog with accept)
source	string	Conditional	CSS selector of element to drag (for drag)
target	string	Conditional	CSS selector of drop target (for drag)
return_snapshot	boolean	No	Return page snapshot (default: true)
include_debug	boolean	No	Include console errors and failed network requests

// List all tabs
{ "action": "tabs", "tab_action": "list" }

// Open new tab
{ "action": "tabs", "tab_action": "create", "url": "https://myapp.example.com/settings" }

// Switch to second tab
{ "action": "tabs", "tab_action": "switch", "index": 1 }

// Resize viewport (mobile)
{ "action": "resize", "width": 375, "height": 812 }

// Execute JavaScript
{ "action": "evaluate", "expression": "document.title" }

// Accept a dialog
{ "action": "handle_dialog", "dialog_action": "accept" }

// Accept a prompt dialog with text
{ "action": "handle_dialog", "dialog_action": "accept", "prompt_text": "My answer" }

// Drag and drop
{ "action": "drag", "source": ".draggable-item", "target": ".drop-zone" }

Caution

The evaluate expression runs in the browser context, not in Node.js. It has access to the DOM and browser APIs but not to server-side variables. Dialogs are auto-captured when they appear, with a 30-second auto-dismiss timeout.