Agent Runtime
Status: Connected
❯Parsing documentation...
✓Extracted api references
❯Generating type definitions...
opensteeropensteer
Y Combinator logoBacked by Y Combinator.

The most comprehensive browser automation framework for AI

get startedgithub

Enterprise-grade automation at scale.

Custom plans, unlimited concurrent sessions, dedicated proxies, and advanced support for teams that need more.

+ self-service sso
+ unlimited concurrent sessions
+ custom audit logs
+ advanced captcha bypass
+ dedicated support
view pricing

Simple, transparent pricing.

Start for free, upgrade when you need to scale. Everything you need to build robust AI agents.

SDK Reference

Navigation and Observation

Navigate pages, read state, capture snapshots/screenshots, and wait reliably.

Raw Playwright

  • page: Page -- the active Playwright Page
  • context: BrowserContext -- the active Playwright BrowserContext

These throw if accessed before launch() or Opensteer.from().

goto(url: string, options?: GotoOptions): Promise<void>

Navigate to a URL and wait for visual stability. Prefer this over opensteer.page.goto() for consistent post-navigation settling.

await opensteer.goto('https://example.com')
await opensteer.goto('https://example.com', { waitUntil: 'networkidle', settleMs: 1500 })

Observation

snapshot(options?: SnapshotOptions): Promise<string>

Return a cleaned HTML snapshot of the current page. Elements are annotated with c="..." counters that can be used in subsequent actions via the element option. Counters are reassigned on each snapshot and synced to the live DOM.

const html = await opensteer.snapshot()
const html = await opensteer.snapshot({ mode: 'extraction' })
state(): Promise<StateResult>

Return the current page URL, title, and an action-mode snapshot.

const { url, title, html } = await opensteer.state()
screenshot(options?: ScreenshotOptions): Promise<Buffer>

Take a screenshot of the current page.

const buffer = await opensteer.screenshot()
const jpeg = await opensteer.screenshot({ type: 'jpeg', quality: 80, fullPage: true })

Cursor Preview

Cursor preview is compositor-based (CDP Overlay), does not inject DOM into the target page, and is best-effort (graceful no-op when unsupported).

  • SDK default: cursor.enabled = false
  • CLI default: enabled for opened sessions
setCursorEnabled(enabled: boolean): void

Enable or disable cursor preview at runtime for this Opensteer instance.

opensteer.setCursorEnabled(true)
getCursorState(): { enabled: boolean; active: boolean; reason?: string }

Return current cursor preview state.

const state = opensteer.getCursorState()

Wait

waitForText(text: string, options?: { timeout?: number }): Promise<void>

Wait for text to appear on the page. Default timeout is 30 seconds.

await opensteer.waitForText('Welcome back')
await opensteer.waitForText('Order confirmed', { timeout: 10000 })

Covered Features

  • sdk:context
  • sdk:getCursorState
  • sdk:getHtml
  • sdk:getTitle
  • sdk:goto
  • sdk:page
  • sdk:screenshot
  • sdk:setCursorEnabled
  • sdk:snapshot
  • sdk:state
  • sdk:waitForText