Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errorsBuild the agent as two cooperating layers: a reasoning model that proposes browser actions and an isolated browser worker that executes those actions, captures a screenshot, and returns evidence. Keep one browser session alive for the whole task, validate every action and navigation, and return a structured result containing status, final URL, screenshot bytes, viewport, and an error trace. Playwright or Selenium can provide the browser runtime; the screenshot itself is your application’s responsibility.
The architecture: model, browser, and evidence
An AI agent that screenshots a site is not just an API call to a language model. It is a controlled loop:
As an Amazon Associate I earn from qualifying purchases.
- Task: accept a URL, an instruction, viewport or device settings, timeout, and authentication policy.
- Reasoning: the model turns the instruction into a small set of browser actions, such as navigate, click a consent button, wait for a selector, or scroll.
- Execution: your service validates those actions and runs them in a real browser session.
- Observation: after each action batch, send the model a screenshot and structured state such as the current URL, visible text summary, and action result.
- Completion: when the target state is reached, capture the final image and return machine-readable evidence to the caller.
This separation matters. The model may suggest an unsafe navigation or misread page content, but the application remains the authority for permissions, origin allowlists, credentials, timeouts, and data handling. Website content is untrusted input and cannot grant itself access to another origin or override your application instructions.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Define a contract your agent can verify
Use a stable request and response shape before choosing a browser provider. A practical request contains:
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
url: the initial URL to open.task: a constrained instruction, for example “open the page, dismiss the consent dialog if present, wait for the article heading, and capture the full page.”viewport: width, height, and optional device or retina scale.timeout_ms: a hard upper bound for navigation and actions.auth_policy: whether existing cookies, custom headers, or an authenticated context may be used.allowed_origins: hosts the worker may visit.
Return at least:
status(for exampleokorerror).final_urlafter redirects and navigation.screenshotas an image payload or a reference to securely stored bytes.viewport, capture timestamp, and browser metadata.actionswith each requested action, validation result, and execution result.errorwith a stable code, human-readable message, and retryability flag.
Returning state with the image lets a reviewer reproduce a result and lets downstream automation distinguish a real page capture from a blank page, timeout, or policy rejection.
Choose the browser runtime
| Decision | OpenAI-hosted browser session | Self-managed Playwright or Selenium |
|---|---|---|
| Control | The service manages the browser environment, session events, and lifecycle. | You control the browser process, image, launch flags, context, and cleanup. |
| Operational work | Create a session, handle events and origin approvals, review activity, then delete the session. | Provide isolation, patch browsers, manage concurrency, retries, storage, and process cleanup. |
| Network policy | Apply the hosted session’s origin-approval flow. | Enforce your own navigation allowlist, DNS and egress policy, and redirect checks. |
| Authentication | Keep the hosted session alive across turns so login state remains available. | Keep one Playwright context alive across turns and explicitly control cookies, headers, and secrets. |
| Latency and scaling | Startup and capacity depend on the hosted service. | You choose warm pools, concurrency, browser reuse, and regional placement. |
| Data residency | Subject to the hosted service’s available regions and policies. | Place workers and storage in regions you operate. |
| Pricing and quotas | Exact figures depend on the service and are not specified here. | Your costs are infrastructure, browser capacity, bandwidth, storage, and engineering time. |
For either pattern, preserve runtime state between model turns. Sending another API message does not recreate a browser’s cookies, local storage, open tabs, or JavaScript variables.
A self-managed implementation with Playwright
Install an isolated worker
The following service exposes a small capture API. It accepts a validated action list from your reasoning layer, keeps one browser context for the request, waits for the requested state, and returns a base64-encoded PNG. Run it in a container or dedicated worker account rather than on a developer workstation.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →python -m venv .venv
. .venv/bin/activate
pip install fastapi uvicorn playwright pydantic
playwright install chromiumDisable extensions, avoid mounting host directories, do not inherit unnecessary environment variables, and apply a network policy at the container or firewall layer. Keep credentials outside logs and never log raw image bytes.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
Complete Python service
import base64
import time
from typing import Literal, Optional
from urllib.parse import urlparse
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel, Field
from playwright.async_api import async_playwright, TimeoutError as PlaywrightTimeoutError
app = FastAPI()
class Viewport(BaseModel):
width: int = Field(default=1440, ge=320, le=3840)
height: int = Field(default=900, ge=240, le=2160)
device_scale_factor: float = Field(default=1.0, ge=1.0, le=3.0)
class Action(BaseModel):
kind: Literal["click", "wait_for_selector", "wait_ms", "scroll"]
selector: Optional[str] = None
milliseconds: Optional[int] = Field(default=None, ge=0, le=30000)
pixels: Optional[int] = Field(default=None, ge=-20000, le=20000)
class CaptureRequest(BaseModel):
url: str
task: str
actions: list[Action] = []
viewport: Viewport = Viewport()
timeout_ms: int = Field(default=30000, ge=1000, le=120000)
allowed_origins: list[str] = []
full_page: bool = True
def origin(url: str) -> str:
parsed = urlparse(url)
if parsed.scheme not in {"http", "https"} or not parsed.hostname:
raise ValueError("Only http and https URLs are allowed")
return f"{parsed.scheme}://{parsed.hostname.lower()}"
def allowed(url: str, origins: list[str]) -> bool:
return not origins or origin(url) in {origin(item) for item in origins}
@app.post("/capture")
async def capture(request: CaptureRequest):
try:
initial_origin = origin(request.url)
except ValueError as exc:
raise HTTPException(status_code=400, detail=str(exc))
if request.allowed_origins and initial_origin not in {origin(item) for item in request.allowed_origins}:
raise HTTPException(status_code=403, detail="Initial origin is not allowed")
started = time.time()
trace = []
try:
async with async_playwright() as pw:
browser = await pw.chromium.launch(headless=True)
context = await browser.new_context(
viewport={"width": request.viewport.width, "height": request.viewport.height},
device_scale_factor=request.viewport.device_scale_factor,
java_script_enabled=True,
)
page = await context.new_page()
await page.goto(request.url, wait_until="domcontentloaded", timeout=request.timeout_ms)
if not allowed(page.url, request.allowed_origins):
raise PermissionError("Redirected to an origin outside the allowlist")
trace.append({"kind": "navigate", "url": page.url, "ok": True})
for action in request.actions:
if action.kind == "click":
if not action.selector:
raise ValueError("click requires selector")
await page.locator(action.selector).first.click(timeout=request.timeout_ms)
elif action.kind == "wait_for_selector":
if not action.selector:
raise ValueError("wait_for_selector requires selector")
await page.wait_for_selector(action.selector, timeout=request.timeout_ms)
elif action.kind == "wait_ms":
await page.wait_for_timeout(action.milliseconds or 0)
elif action.kind == "scroll":
await page.evaluate("pixels => window.scrollBy(0, pixels)", action.pixels or 0)
trace.append(action.model_dump() | {"ok": True})
image = await page.screenshot(type="png", full_page=request.full_page)
result = {
"status": "ok",
"final_url": page.url,
"screenshot": {
"media_type": "image/png",
"base64": base64.b64encode(image).decode("ascii")
},
"viewport": request.viewport.model_dump(),
"elapsed_ms": round((time.time() - started) * 1000),
"actions": trace,
"task": request.task,
}
await context.close()
await browser.close()
return result
except PlaywrightTimeoutError as exc:
return {"status": "error", "error": {"code": "timeout", "message": str(exc), "retryable": True}, "actions": trace}
except PermissionError as exc:
return {"status": "error", "error": {"code": "origin_not_allowed", "message": str(exc), "retryable": False}, "actions": trace}
except Exception as exc:
return {"status": "error", "error": {"code": "capture_failed", "message": str(exc), "retryable": False}, "actions": trace}
Start it with:
uvicorn capture_service:app --host 0.0.0.0 --port 8000The model should emit only actions your schema accepts. Your service can add a second validation pass for selector length, JavaScript URLs, maximum scroll distance, and navigation destinations. The example deliberately does not execute arbitrary model-supplied JavaScript.
Call the service with cURL
curl -X POST http://localhost:8000/capture
-H 'content-type: application/json'
-d '{
"url":"https://example.com",
"task":"Open the page and capture the full document",
"actions":[{"kind":"wait_for_selector","selector":"h1"}],
"allowed_origins":["https://example.com"]
}'Connect the reasoning model safely
Give the model a narrow task and a tool schema rather than unrestricted browser access. A useful turn contains the current URL, viewport, a screenshot or accessibility observation, and the actions already attempted. Ask for one short action batch at a time. After execution, send back:
- the updated screenshot;
- the final URL and whether it passed the origin policy;
- which selectors were found or missing;
- navigation, console, and network errors in structured form;
- the remaining time and retry budget.
Stop when the required selector is visible and the page is stable, or when the model reaches a bounded action count. The application—not page text—decides whether a click, upload, download, cross-origin navigation, or credential use is permitted. For hosted browser sessions, monitor session events and answer origin-access approval requests explicitly. For self-managed sessions, perform the equivalent check before every navigation.
Full-page, element, and authenticated captures
Waiting for dynamic content
Prefer a meaningful readiness condition, such as a main-content selector, over a fixed sleep. Use a short delay only for animations or late layout shifts, and cap it. If a site loads images lazily, scroll in bounded increments before the final full-page screenshot and verify that the expected image or content count is present.
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
Capturing one element
Replace the final page screenshot with a locator screenshot when the contract asks for a component, chart, or invoice rather than the whole document. Validate the selector first and return the element’s bounding box in the evidence record so a reviewer knows what was captured.
Login and secrets
Create a dedicated browser context for each tenant or job. Supply cookies or headers through a secret manager, never through the task text. Do not persist a profile unless you have a specific retention policy. Screenshots can contain account and personal data, so encrypt stored images, restrict access, set retention limits, and avoid logging image payloads.
Hosted browser sessions: approvals and continuity
An OpenAI-hosted browser pattern moves browser provisioning and much of the lifecycle work to the service. Your code still creates a session, submits turns, consumes session events, handles origin approvals, reviews browser activity, and deletes the session when finished. Keep the same session for all turns in one task; a new conversation message alone does not restore login state or runtime variables.
This option reduces browser maintenance but does not remove policy work. You still need a task allowlist, credential policy, timeout and retry limits, screenshot-size limits, secure storage, and a decision about which origins the agent may visit. Exact hosted pricing, quotas, and regional availability are not established here, so check the current service documentation before committing capacity.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
Reliability, performance, and cost controls
- Warm browsers: reuse a process or pool when safe, but create isolated contexts so cookies and local storage do not leak between jobs.
- Bound every wait: apply navigation, selector, action-batch, and total-job deadlines.
- Retry selectively: retry transient network failures and browser crashes with a fresh context; do not retry policy denials or invalid selectors indefinitely.
- Cap output: enforce maximum pixel dimensions and encoded byte size. Downscale or reject oversized full-page images before they enter queues or logs.
- Record timing: measure browser startup, navigation, readiness wait, action execution, screenshot encoding, and upload separately.
- Control concurrency: limit pages per worker and account for CPU, memory, file descriptors, bandwidth, and downstream model-image limits.
- Cache deliberately: cache only when URL, headers, cookies, viewport, device scale, task state, and freshness requirements match. Never serve a private capture from a public cache.
- Budget model turns: provide the model with a compact observation and stop after a fixed number of unsuccessful attempts.
Self-managed cost is dominated by worker infrastructure, browser capacity, bandwidth, storage, and engineering operations. Hosted cost depends on the provider’s session and usage terms; do not assume a quota or regional price that has not been confirmed.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Troubleshooting common failures
| Symptom | Likely cause | Fix |
|---|---|---|
| Navigation timeout | The origin is slow, blocked, or waiting on a resource that never completes. | Use a total deadline, wait for domcontentloaded plus a target selector, inspect network errors, and retry once in a fresh context. |
| Blank or mostly white image | Capture occurred before rendering, a script failed, or the page requires a blocked resource. | Wait for a meaningful selector, check console and failed requests, verify the viewport, and capture only after the intended state is visible. |
| Consent dialog blocks content | The model did not identify the dialog or the selector changed. | Expose visible buttons as observations, allow a bounded click on an approved selector, then verify the dialog is gone. |
| Redirect rejected | The destination is outside the allowlist. | Record the redirect and require an explicit policy update; never silently broaden the allowlist from page content. |
| Selector not found | Wrong selector, iframe, delayed rendering, or a shadow DOM boundary. | Inspect the DOM and frame tree, wait for a readiness condition, and provide a selector that is stable across renders. |
| Login disappears between turns | A new browser context or process was created. | Keep the same session alive for the task and verify cookie or storage state without exposing secrets to the model. |
| Out-of-memory worker | Too many pages, oversized full-page captures, or leaked browser processes. | Limit concurrency and image dimensions, close contexts in a finally path, and recycle unhealthy workers. |
| Model loops on the same action | No action budget or structured failure feedback. | Return the exact failure code, count attempts, cap turns, and terminate with a reproducible trace. |
Or skip the browser setup
ScreenshotNeo is a hosted website screenshot API and MCP server. It accepts a URL in one request and returns PNG, JPEG, WebP, or PDF. Before capture it can accept cookie or consent banners and remove more than 60 known consent platforms, newsletter popups, and chat widgets; each step can be disabled. Bot checks or CAPTCHAs, blank pages, timeouts, failed loads, and cache hits are not billed, and each response identifies the outcome with X-Page-Verdict and X-Billed headers. Its MCP server exposes take_screenshot, get_page_info, and capture_pdf for Claude, Cursor, and other MCP clients.
One call is enough for a basic capture:
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpPython:
import requests
r = requests.get("https://api.screenshotneo.com/v1/shot", params={"access_key": "YOUR_API_KEY", "url": "https://stripe.com"}, timeout=90)
open("shot.webp", "wb").write(r.content)Node.js:
const q = new URLSearchParams({ access_key: 'YOUR_API_KEY', url: 'https://stripe.com' });
const res = await fetch(`https://api.screenshotneo.com/v1/shot?${q}`);
const body = Buffer.from(await res.arrayBuffer());
await import('node:fs/promises').then(fs => fs.writeFile('shot.webp', body));See the parameter reference and integration details in the ScreenshotNeo documentation. The service supports 63 options, including full-page capture with lazy images loaded, CSS-selector element capture, dark mode, 12 device presets plus custom viewports, retina scale, PDF paper size, margins, landscape mode and page ranges, HTML/CSS-to-image, custom CSS and JavaScript, pre-capture clicks, hidden selectors, waits for selectors, delays or network idle, ad and tracker blocking, request or resource-type blocking, custom headers, cookies, user agents and Authorization, timezone and geolocation, transparent backgrounds, image resizing, selectable-TTL caching, signed links for public image tags, asynchronous jobs with signed webhooks, bulk capture of up to 100 URLs per call, a usage API, an OpenAPI specification, and compatibility with parameter names used by other screenshot APIs.
Plans are Free with 1,000 shots per month and no card, Starter at $5 for 3,000, Growth at $15 for 15,000, Pro at $39 for 60,000, Scale at $99 for 250,000, and Business at $249 for 1,000,000; yearly billing gives two months free, and every feature is included on every plan. Create a free ScreenshotNeo account to start with 1,000 screenshots a month and no card.
Security checklist before production
- Run browsers in isolated containers or workers with no unnecessary extensions, host-file mounts, or inherited environment variables.
- Use HTTPS for every control-plane and image-transfer endpoint.
- Allow only approved schemes, hosts, ports, redirects, downloads, and frame origins.
- Keep credentials in a secret manager and redact them from traces, URLs, headers, and model observations.
- Set maximum action count, navigation time, screenshot size, and total job duration.
- Encrypt screenshots at rest, restrict access, and delete them according to a documented retention period.
- Audit every action and approval without storing raw page content unnecessarily.
- Test hostile pages, prompt-injection text, redirects, infinite scroll, popups, downloads, and cross-origin frames.
FAQ
Does the model take the screenshot itself?
No. The model proposes actions and interprets observations; your browser worker or a hosted browser executes the actions and captures the image.
Should I use Playwright or Selenium?
Either can provide the browser runtime. Choose based on your team’s language support, existing automation, isolation model, and operational tooling; the important property is a persistent, policy-controlled session.
Why return a screenshot after intermediate actions?
Intermediate images let the model and your reviewer verify that a click, consent dismissal, redirect, or lazy-load step produced the intended state before the final capture.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Can I let a page choose the next origin?
Not automatically. Treat page content as untrusted and require your application’s allowlist and approval logic for every cross-origin navigation.
Frequently Asked Questions
Does the model take the screenshot itself?
No. The model proposes actions and interprets observations; your browser worker or a hosted browser executes the actions and captures the image.
Should I use Playwright or Selenium?
Either can provide the browser runtime. Choose based on your team’s language support, existing automation, isolation model, and operational tooling; the important property is a persistent, policy-controlled session.
Why return a screenshot after intermediate actions?
Intermediate images let the model and your reviewer verify that a click, consent dismissal, redirect, or lazy-load step produced the intended state before the final capture.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOutdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchCan I let a page choose the next origin?
Not automatically. Treat page content as untrusted and require your application’s allowlist and approval logic for every cross-origin navigation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
