Does safari_wait_for change anything on the page?

No. It is a read-only tool. It only polls your current Safari tab until the CSS selector appears (or its text matches text_match) and then reports back. It never clicks, types, or modifies the page.

How does it know when to stop waiting?

It repeatedly checks the page for the CSS selector you describe. If you also provide a text_match value, it waits until the matched element's text contains that text. Once the condition is met, it returns immediately.

Which AI assistants can use it?

Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed configure automatically via local stdio. Claude.ai, ChatGPT, Grok and Perplexity connect through the LMCP Cloud Relay connector.

Is any of my browsing sent to the cloud?

No. LMCP runs natively on your Mac and reads Safari locally. No page content is stored on a server and no API keys are needed, so it is GDPR-compliant by architecture.

Wait for a Safari Element to Load with AI on Mac

safari_wait_for tells your AI assistant to keep checking your current Safari tab until a specific element shows up. It polls the page for a CSS selector you describe in plain English (and optionally waits until that element's text matches), then returns once the condition is met. This is the missing piece for reliable web automation: instead of guessing how long a page takes to load, the AI waits for the actual signal. A natural example you can type: "In Safari, wait until the order confirmation message appears, then read it back to me." Behind the scenes the AI maps that to safari_wait_for with a selector like .confirmation-banner and an optional text_match such as "Order placed".

It is a read-only tool — it never clicks, types, or changes anything. It just watches and reports when the page is in the state you asked for. Download LMCP to add it to your Mac.

Which AI agents work?

LMCP exposes safari_wait_for to every major AI client through the Model Context Protocol, so you can use whichever assistant you already have open:

Desktop clients (auto-configure via local stdio): Claude Desktop, Cursor, VS Code (GitHub Copilot), Windsurf and Zed. LMCP wires itself into their MCP config automatically.
Web AIs (via the LMCP Cloud Relay connector): Claude.ai, ChatGPT, Grok and Perplexity connect securely to your Mac through the relay so they can drive Safari on your machine.

One install, 150+ tools, available to all of them.

Automation

safari_wait_for is a synchronization primitive — it shines when chained between other Safari tools. A typical flow: open a URL, click a button to submit a form, then safari_wait_for the success element before reading the result. Without it, the AI might try to read or extract content before the page finishes rendering and grab nothing. By inserting a wait step, multi-step browser workflows become dependable. Tell the AI: "Log into the dashboard, wait for the charts to render, then summarize this week's numbers." and it sequences click, wait, and read for you.

Context

This runs against your real Safari session on your Mac — the tabs you already have open, with your logins and cookies intact. The AI sees exactly the page you see, including content behind authentication or single-page apps that load asynchronously. There is no separate headless browser and no need to re-authenticate; the assistant works with your live browsing context, which is what makes the wait condition meaningful in real workflows.

Productivity

Manually babysitting a slow-loading page — refreshing, squinting, waiting for a spinner to disappear — is exactly the kind of small friction that adds up. With safari_wait_for, the AI absorbs that waiting for you and acts the instant the page is ready. Whether you are watching for a deploy status to flip to "green", a long report to finish generating, or a checkout to confirm, you hand off the polling and get notified with the result. It turns brittle, timing-dependent tasks into a single sentence and saves the dead minutes you'd otherwise spend staring at a loading screen.

Privacy & GDPR

Everything happens locally on your Mac. safari_wait_for reads your current tab directly through native macOS automation — no page content is copied to a cloud server, no API keys are required, and nothing is stored remotely. The AI receives only the result of the wait you asked for. Because the architecture keeps your data on your own machine, LMCP is GDPR-compliant by architecture. Your browsing stays yours.