Glimpsa — AI at Your Cursor & Sidebar

AI at your cursor and in your sidebar: select or right-click anything for an instant, context-aware answer. Privacy-first.

✦ AI at your cursor. And in your sidebar. Stop tab-switching to ask an AI about the thing already on your screen. Select any text — or right-click an image, link, table, or email — and Glimpsa streams the answer into a tidy popover right where you're reading. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ⚡ ONE GESTURE. AN ANSWER. Highlight a phrase and a small chip appears with the actions that actually fit what you picked — nothing more. • Explain that dense paragraph in plain English • Summarize a long article down to what matters • Translate into your language, instantly • Identify what's in an image, chart, or screenshot • Extract a table as clean, pasteable data • Draft a reply to the email you're staring at Ask follow-ups in the same popover. Upgrade any answer to a stronger model with one tap when the quick answer isn't enough. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ✓ ANSWERS YOU CAN ACTUALLY CHECK This is the part other assistants don't do. Generic AI generates. Glimpsa verifies. Tap Verify and your answer is re-checked against live web sources — claim by claim — and each one comes back marked: ✓ Supported — found in a real source, linked ⚠ Unverified — nothing solid either way ✗ Contradicted — the sources disagree Because "sounds confident" and "is correct" are not the same thing, and you're the one who has to act on it. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ▤ A SIDEBAR THAT READS THE WHOLE THING Click the toolbar icon to dock the assistant, then: • Chat about the page you're on — summarize it, pull the key points • Drop in a screenshot or a scan and ask what's in it • Attach a PDF and interrogate it — text is extracted on your machine • Draft an email from any of it ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ✉ BUILT FOR YOUR INBOX On Gmail and Outlook, Glimpsa reads the open message and drafts a reply in your tone — shorter, friendlier, firmer, more formal, your call — then drops it straight into the compose box. No copy-paste. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ⛨ PRIVACY-FIRST, AND WE MEAN THE ARCHITECTURE Nothing leaves your machine until you tap an action. Selecting text and hovering are entirely local — zero network calls, by design. Bring your own Anthropic API key and your content goes straight from your browser to Anthropic. There is no backend in the loop. Not a policy — there is simply no server to store anything on. Don't want to manage a key? Sign in and we relay your request to Anthropic for you. Even then we count requests, never content: your prompts and answers are never stored. Calm by default. The chip appears on selection or right-click — never unprompted. Disable it on any site you like. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ◈ NO LOCK-IN Built on Claude, running on your key if you want it to. Not wired into any one browser vendor's ecosystem, not harvesting your reading habits to sell you something. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ▸ PRICING FREE • 3 answers to try — no sign-up, no key, nothing to configure • Sign in for 3 free every day • Or bring your own Anthropic key: unlimited and fully private PRO — $9.99/month • The Verify layer + "Go deeper" on a stronger model • Gmail & Outlook reply insert • PDF and document chat • Your own custom prompts as one-tap actions • Works two ways: our servers with a generous monthly quota and no key to manage, or your own key for unlimited private use Cancel anytime. 14-day refund, no questions asked. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Install it, highlight a sentence, and see. Built on Claude. Not affiliated with Anthropic or Google.