Docs/Features/Browser Control
Features

Browser Control

Drive your real Chrome browser β€” your actual logged-in tabs and sessions β€” through a companion Chrome extension, with a plan-and-evidence layer to keep multi-step tasks honest.

Two tools power this: browser_tab for the actions, and browser_plan for keeping multi-step browser tasks honest.

Both tools are off by default. They register only when

tools.browserTab.enabled is true in ~/.flowly/config.json; browser_plan registers alongside browser_tab.

browser_tab

Drives your real Chrome through the Flowly extension. A single action parameter selects what to do, with per-action arguments. The available actions:

ActionPurpose
read_pageRead the current page
get_page_textExtract page text
navigateGo to a URL
clickClick an element
typeType text
form_inputFill a form field
upload_fileUpload a file
upload_imageUpload an image
hoverHover over an element
findFind an element
waitWait
evaluateEvaluate JavaScript
console_logRead console logs
dialogHandle a dialog
screenshotCapture the page
scrollScroll
keySend a keystroke
tabs_listList open tabs
tabs_createOpen a new tab
tabs_closeClose a tab
tabs_contextTab context
read_network_requestsRead network requests
batchRun a batch of actions

Because it operates on your real browser, the agent works inside whatever sites you are already signed in to β€” no separate login or headless session.

browser_plan

An explicit plan-and-evidence layer for browser tasks. Long browser flows tend to drift after many tool calls β€” the agent forgets what it was doing or reports success when the page state does not match the request. browser_plan solves this by making the agent commit to a plan up front and verify each step with evidence before marking it done.

It has four actions:

ActionPurpose
createStart a plan (goal, steps)
viewView the current plan
update_stepUpdate a step's status (must attach evidence before done)
completeComplete the plan

Each step carries a successCriteria and an evidence slot β€” a screenshot description, DOM observation, or URL change β€” that must be filled before the step can move to done.

How the connection works

browser_tab talks to the Flowly Chrome extension, which controls your real browser tabs. You enable the tool and confirm the extension is connected through the /browser modal (below).

Until the extension is installed and connected,

browser_tab actions cannot reach the browser.

The /browser slash command

In the TUI, /browser opens the Browser Use modal, where you can:

  • Toggle the browser_tab enable flag on or off.
  • See live extension-connection status.
  • Open the Chrome Web Store to install the extension if it is not installed yet.

Saving the toggle restarts the gateway so the change takes effect; the TUI confirms with browser_tab enabled / disabled Β· gateway restarted.

Configuration

KeyPurpose
tools.browserTab.enabledEnables browser_tab (and browser_plan). Default off.
You can flip this key by hand in

~/.flowly/config.json, but using /browser is preferred since it also restarts the gateway for you.