Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Use an MCP-capable client, Playwright MCP, and a local model provider such as Ollama. The model does not drive Playwright directly: the client connects to the Playwright MCP server, exposes browser tools to the model, and returns structured accessibility snapshots and element references. The reliable loop is navigate → inspect → act → inspect again. Node.js 20 or newer and an MCP-capable client are Playwright’s documented prerequisites. A local model must also support tool use in the exact client you choose; installing each component separately does not guarantee compatibility.
What the architecture looks like
Model choice, client integration and browser automation are separate layers:
- Local model: Ollama runs a model on your computer and exposes it to supported clients.
- MCP client: An application such as VS Code Chat mediates between the model and MCP tools.
- Playwright MCP: The server launches a browser and exposes navigation, inspection and interaction tools.
- Browser: Playwright performs the actual page operations.
Playwright describes its server this way: “The Playwright MCP server provides browser automation capabilities through the Model Context Protocol, enabling LLMs to interact with web pages using structured accessibility snapshots.” Ordinary snapshot-based interaction does not require a vision model.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Prerequisites
- Node.js 20 or newer.
- An MCP-capable client. The examples below use VS Code Chat.
- Ollama installed and running locally.
- A local model available to Ollama that can follow tool calls in your client.
- A test website that you are authorized to automate.
These requirements can change as clients and extensions evolve. Check the current Playwright MCP, VS Code and Ollama documentation when setting up a new machine.
#1 Best Overall
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
Set up a local model with Ollama in VS Code
1. Install Ollama and obtain a model
Install Ollama for your operating system, start its local service, and make a model available according to Ollama’s current instructions. Ollama’s VS Code integration discovers models from http://127.0.0.1:11434 by default. Local models do not require an Ollama sign-in for this workflow.
2. Install the VS Code extension
Install the Ollama extension in VS Code, open Chat, and select the local model from the model picker. If no model appears, confirm that the Ollama service is running, that the extension is using the expected local endpoint, and that the model is installed locally.
3. Confirm tool-calling support
Send a harmless normal-chat request first, then test a small browser task after MCP is connected. A model can generate text successfully yet fail to emit the structured tool calls required by Playwright. There is no general compatibility guarantee across every Ollama model, VS Code version and MCP client.
Add Playwright MCP
Playwright’s standard MCP configuration launches the server with npx:
{
"mcpServers": {
"playwright": {
"command": "npx",
"args": ["@playwright/mcp@latest"]
}
}
}Add this server through your client’s MCP-server workflow. VS Code can configure MCP servers from its documented MCP interface; the exact file location and UI labels may change, so use the current client instructions rather than copying an old path.
When the client starts the server, npx downloads or uses the requested Playwright MCP package and the server exposes browser tools. Keep Node.js 20 or newer on the system where the client starts it.
Run the first browser task
- Open VS Code Chat with the Ollama model selected.
- Ask the assistant to open https://demo.playwright.dev/todomvc.
- Ask it to inspect the accessibility snapshot and identify the input used to add a todo.
- Ask it to fill that control with a benign value and submit it.
- Ask it to inspect the updated page and confirm the new item.
Be explicit that the assistant should inspect before acting. Playwright MCP returns references from the current snapshot; after navigation or a page change, previously returned references may no longer be valid. A good prompt is:
Free tools Windows power users keep installed
One-click scans. No signup required.
Open https://demo.playwright.dev/todomvc. Inspect the accessibility snapshot first. Add one todo named "MCP smoke test", then inspect the page again and report the visible todo items. Do not delete or modify anything else.Understand the inspect-and-act loop
Navigate
Give the assistant a URL and a narrowly scoped objective. For production sites, include authentication, data-handling and allowed-action boundaries in the prompt.
Rank #2
- EVOLUTION AMD RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 64GB pool, which is perfect for running LLMs such as Deepseek 32B, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 4% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Inspect
The model reads the current accessibility snapshot, including roles, names and references for actionable elements. This is generally more robust than guessing CSS selectors from page text.
Interact
The model uses the returned reference to click, fill, select or otherwise operate the control. Ask for one logical action at a time when a workflow is sensitive.
Inspect again
After every navigation, submission, modal change or asynchronous update, request a fresh snapshot. The second inspection verifies what actually happened instead of relying on the model’s assumption.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Choose Playwright MCP capabilities deliberately
Core browser automation is available by default. Playwright also documents optional capability groups:
| Capability group | Use it when you need | Trade-off |
|---|---|---|
| Network | Request interception or network mocking | More tools and schemas for the model to choose from |
| Storage | Cookies, local storage or saved authentication state | Persistent state can expose sensitive data if shared |
| Testing | Test-oriented browser workflows | Additional tool surface beyond a simple task |
| Vision | Tasks that genuinely require visual understanding | Not needed for ordinary accessibility-snapshot interaction |
| PDF generation or inspection workflows | Extra context and tools | |
| DevTools | Debugging browser behavior | Broader access than a basic automation task |
Expose only the groups your workflow needs. Fewer tools reduce schema size and the number of choices presented to the model, which can make tool selection easier.
Persistent versus isolated browser state
Persistent profiles
Playwright uses a persistent browser profile by default. Cookies, local storage and other profile data can remain between tasks. This is useful when a workflow needs a logged-in session, but it also means a later prompt may operate with earlier credentials or site state.
Isolated sessions
Use an isolated context or storage-state strategy when each run must start cleanly. This is preferable for repeatable tests, multi-user scenarios and work involving untrusted pages. Decide deliberately whether authentication state should be reused, and keep saved state out of source control.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Security and trust boundaries
Review MCP configuration before trusting a repository. Microsoft warns: “Review workspace MCP configuration before you trust a repository because local MCP servers can run code on your machine.” A server started by your client has the permissions of that local process.
Rank #3
- Intel Core Ultra 9 285 Processor: Newly developed cores deliver ultra-smooth and responsive gameplay. AI accelerators prepare users for the next era of gaming on an AI PC.
- Simplistic Design: Enjoy the latest generation of Windows 11 Home for your everyday needs. *MSI recommends Windows 11 Pro for business use.
- NVIDIA GeForce RTX 5070 Ti GPU
- Cool While Gaming: In conjunction with an RGB CPU Air Cooler, the Aegis RS features four system cooling fans; three in the front and one in the rear to pull in cool air and push heat out of the PC.
- Turn on the Bright Lights: With the built-in RGB lighting, take your gaming experience to the next level by pressing the MSI LED button to cycle through lighting options. Customize lighting even further with MSI Center software.
Playwright identifies browser_run_code_unsafe as equivalent to remote-code execution because it can execute arbitrary JavaScript in the server process. Do not expose that capability to an untrusted model or client. Prefer the narrowest tool set, use a dedicated account and browser profile, avoid real customer data, and inspect workspace configuration before enabling a server.
Common failures and fixes
The MCP server will not start
- Cause: Node.js is older than 20 or unavailable on the client’s PATH.
- Fix: Install Node.js 20 or newer, restart the client, and verify that the same user environment can run
npx.
The client shows no Playwright tools
- Cause: Invalid JSON, the server was added to the wrong MCP configuration, or the client has not reloaded it.
- Fix: Validate the JSON, confirm the server name and command, reload the client, and inspect its MCP connection log.
Ollama is installed but no model appears
- Cause: Ollama is not running, the model is not installed, or the extension is pointed at a different endpoint.
- Fix: Start Ollama, make a model available locally, and check the configured endpoint, normally
http://127.0.0.1:11434.
The model chats but never calls a tool
- Cause: The selected model or client does not support the required tool-calling format.
- Fix: Try a model documented as tool-capable by your client, update the client and extension, and test with a tiny navigation request.
An element reference stops working
- Cause: The page changed after the snapshot was taken.
- Fix: Ask for a fresh accessibility snapshot, then perform the action using the new reference.
The browser is logged out or data differs between runs
- Cause: You are using an isolated context, or a persistent profile was replaced or cleared.
- Fix: Choose persistent or isolated state intentionally and verify the profile and storage-state settings.
A page appears visually obvious but the model cannot find a control
- Cause: The required information is not exposed accessibly, or the task is genuinely visual.
- Fix: Improve the page’s accessible names where you control it, or enable the vision capability and use a model/client combination that supports it. Do not assume vision is required for ordinary pages.
Performance, reliability and cost considerations
Local inference avoids sending prompts and page content to a hosted model, but speed depends on your hardware, model size, context length and the amount of snapshot data. Start with a small task and measure your own workflow; the available documentation does not establish a general speed or reliability benchmark.
Keep prompts narrow, request snapshots only when state changes, and enable only needed capability groups. For repeatable automation, use deterministic test data, isolated browser state and explicit post-action checks. For destructive actions, require a human confirmation step in the prompt or client workflow.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesPlaywright MCP itself does not remove a website’s consent banners, popups or chat widgets. If your goal is a clean image rather than interactive browser control, a screenshot API can be simpler.
Or skip the browser setup
ScreenshotNeo returns a website screenshot or PDF from one request, so you do not need to configure Playwright, a browser profile or an MCP client for a capture-only job. Before the capture it accepts cookie and consent banners and removes more than 60 known consent platforms, newsletter popups and chat widgets; each cleanup step can be turned off. Bot checks, CAPTCHAs, blank pages, timeouts, failed loads and cache hits are not billed as clean shots, and the response identifies the result with X-Page-Verdict and X-Billed headers.
For API details, see the ScreenshotNeo documentation. cURL:
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}`);ScreenshotNeo also provides an MCP server with take_screenshot, get_page_info and capture_pdf tools for Claude, Cursor and other MCP clients. Every plan includes its features: 1,000 shots per month are free with no card; paid plans start at $5 for 3,000 shots. Sign up for the free plan.
FAQ
Frequently Asked Questions
Can I use any Ollama model with Playwright MCP?
No. The model, client and MCP integration must support the same structured tool-calling behavior. Verify the exact combination with a small safe task.
Does Playwright MCP require Chromium?
The documented setup starts the Playwright MCP server through npx; use the browser installation and browser support requirements specified by the current Playwright documentation for your environment.
Is local browser automation automatically safe because the model is local?
No. The MCP server and any unsafe code-execution tool still run with local process permissions, so treat configuration, pages and prompts as trust-sensitive.
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.
