Questions, answered.

ClipAction is a Chrome extension that turns any selected text on the web into an AI-powered action. Select text, press Ctrl+Shift+Space, pick an action (rewrite, summarize, translate, explain, fix grammar, etc.), and paste the result back — all keyboard-driven.

1. Select any text on a web page.
2. Press Ctrl+Shift+Space (or ⌘+Shift+Space on Mac).
3. Arrow-key or type to filter actions, press Enter to run.
4. Review the result. Press Enter again to copy-and-replace, or Esc to dismiss.

No mouse required. See the setup guide for a walkthrough.

Any Chromium-based browser: Chrome, Brave, Edge, Arc, Opera, Vivaldi.

On-device Gemini Nano requires Chrome 127+ (or a recent Chromium build that ships Nano). Cloud providers work in every supported browser without extra setup.

Yes. The core actions (rewrite, summarize, translate, explain, fix grammar, shorten, expand) are free and run on-device via Gemini Nano. Pro ($5/month) adds custom actions, site profiles, and cloud provider support. See the pricing section.

Gemini Nano is a small language model that ships inside Chrome itself. It runs entirely on your machine — CPU, GPU, or NPU depending on hardware — and never sends your text anywhere. ClipAction uses the Chrome Prompt API, Rewriter API, and Summarizer API to talk to it.

• OS: Windows 10/11, macOS 13+, Linux, or ChromeOS
• Browser: Chrome 127 or newer (check via chrome://version)
• Storage: ~22 GB free for the model download
• GPU: 4+ GB VRAM recommended (integrated GPUs work, but slower)
• Memory: 8 GB RAM minimum
• Network: Unmetered connection for the initial model download (~1.5–2 GB)

No Google account required — the model is downloaded and run locally.

Chrome ships Nano behind two flags. Enable them, restart, and the model downloads in the background.

1. Open a new tab and navigate to chrome://flags.
2. Search for "Prompt API for Gemini Nano" and set it to Enabled.
3. Search for "Enables optimization guide on device" and set it to Enabled BypassPerfRequirement.
4. Click Relaunch at the bottom of the page.
5. After restart, open chrome://components.
6. Find Optimization Guide On Device Model and click Check for update. The version should change from 0.0.0.0 to something like 2024.x.x.x.
7. Wait for the download to complete (can take 5–20 minutes depending on connection).

Once the component version is non-zero, open ClipAction's popup — the provider badge should show ● On-device.

Common causes:

• Not enough free disk space. Chrome needs ~22 GB free on the drive Chrome is installed to. Clear space and click Check for update again.
• Metered connection. Chrome refuses to download large models over metered Wi-Fi. Disable metered mode in system settings or switch networks.
• Enterprise-managed Chrome. If Chrome is managed by your org, the admin may have disabled Nano. Check chrome://policy.
• Flag didn't stick. Re-open chrome://flags and confirm both flags still show Enabled. Relaunch once more if they reset.
• Wait longer. The first download can take 15+ minutes. Leave Chrome open.

Two quick checks:

1. Open chrome://on-device-internals (Chrome 128+) — it shows model status, download progress, and memory usage.
2. Open DevTools on any page, run await ai.languageModel.availability(). If it returns "available", you're good.

Inside ClipAction, the popup shows a green On-device badge when Nano is ready. The extension also runs a quick capability probe on first use.

Try in order:

1. Fully quit Chrome (not just the window — check the tray / menu bar) and reopen.
2. Visit chrome://components → click Check for update on Optimization Guide On Device Model.
3. If the version is still 0.0.0.0, your machine may not meet the hardware bar. Fall back to a cloud provider (Pro) — Groq and Gemini 2.0 Flash are fast and cheap.
4. Open chrome://on-device-internals for a detailed error. Share that with support if you're stuck.

Yes, after the initial model download. Once Nano is installed, you can run ClipAction actions with Wi-Fi off.

Nine: OpenAI, Anthropic, Google Gemini, Groq, Mistral, xAI, Together, OpenRouter, and Ollama (local-network). Cloud providers require Pro and your own API key.

• OpenAI: platform.openai.com/api-keys
• Anthropic: console.anthropic.com/settings/keys
• Google Gemini: aistudio.google.com/apikey (free tier available)
• Groq: console.groq.com/keys (free tier available)
• Mistral: console.mistral.ai/api-keys
• xAI: console.x.ai
• Together: api.together.xyz/settings/api-keys
• OpenRouter: openrouter.ai/keys
• Ollama: No key — runs locally on localhost:11434.

Open the ClipAction popup → click Settings in the footer → AI Providers tab. Expand the provider you want, paste the key, and hit Test. Pick your default via the Model Routing tab. See the Intelligence Center.

No. Keys are stored in Chrome's local encrypted storage (chrome.storage.local) on your device. They're used only to authenticate requests that go directly from your browser to the provider's API.

We don't run any servers that see your keys or your text.

Rough guide:

• Free & fast: Gemini Nano (on-device) for anything short — rewrites, grammar fixes, summaries under ~500 words.
• Speed matters: Groq (Llama 3.3 70B) — often 300+ tokens/sec.
• Long context / nuance: Claude 3.5 Sonnet or GPT-4o.
• Cheap cloud: Gemini 2.0 Flash ($0.10/M in) or Mistral Small.
• Fully local with a bigger model: Ollama on your own machine.

1. Install Ollama.
2. Pull a model, e.g.

   ollama pull llama3.2

3. Make sure Ollama is running on localhost:11434.
4. In ClipAction → Settings → AI Providers, expand Ollama and pick the model you pulled.
5. ClipAction will request host permission for http://localhost:11434/* the first time — approve it.

None. No analytics, no telemetry, no account, no user ID. See the full privacy policy.

By us: no — we never see your text.

By providers: depends on the provider. On-device Nano runs locally and trains nothing. Cloud APIs (OpenAI, Anthropic, etc.) have their own data policies — most don't train on API traffic by default, but verify with your provider's enterprise settings if you're handling sensitive content.

● On-device means the action ran via Gemini Nano or Ollama — your text stayed on your machine.

● Cloud means the action was sent to a remote provider's API. You'll see this any time you pick a cloud model.

Every result shows one of these two badges so you always know where your text went.

Only when you invoke it. The extension uses the activeTab permission — it has no access to a tab until you press the shortcut or click its icon on that tab. It does not observe your browsing, keystrokes, or clipboard.

Technically yes, but treat it like any other AI tool: if your company has a policy against pasting data into third-party AI, stick to on-device Nano or Ollama — neither leaves your machine. Cloud providers see whatever text you send them.

The core actions — rewrite, summarize, translate, explain, fix grammar, shorten, expand — all running on-device via Gemini Nano. No API keys, no cost, no account. The free tier is enough to decide whether ClipAction fits how you work.

• Cloud providers — bring your own API keys for OpenAI, Anthropic, Google, Groq, Mistral, xAI, Together, OpenRouter
• Ollama — connect a local Ollama server for bigger on-device models
• Custom Prompts — author your own actions with {{selection}}, {{language}}, {{tone}}, {{length}} variables
• Site Profiles — per-domain action palettes (GitHub, Gmail, Slack, Linear, custom)
• Chains — run actions in sequence, feeding one result into the next
• Custom shortcuts — bind keyboard combos to individual actions

$5/month.

From the ClipAction popup, click the Upgrade button at the top. If you've already paid on another machine, use Already Pro? Restore purchase just below it.

No trial. The free plan already gives you the core actions running on-device — that's the best way to decide if ClipAction fits how you work before upgrading.

Cancellation is handled through the billing portal you used at checkout. Cancel there any time — your subscription stays active until the end of the current billing period, then the extension falls back to free-tier behaviour. If you can't find the link, email support and we'll sort it.

Ctrl+Shift+Space on Windows/Linux, ⌘+Shift+Space on Mac.

1. Open chrome://extensions/shortcuts.
2. Find ClipAction.
3. Click the pencil icon next to Open ClipAction command palette.
4. Press your preferred key combination, then click OK.

Chrome reserves some combinations (anything system-wide like Ctrl+T). If your chosen keys are taken, Chrome will show a conflict warning.

Checks in order:

1. Visit chrome://extensions/shortcuts and confirm the shortcut is actually assigned.
2. Make sure you've selected some text on the page first — no selection, no palette.
3. Reload the tab. The content script doesn't inject into tabs that were open before the extension was installed.
4. Check chrome://extensions → ClipAction is Enabled and Allowed in incognito if you're in a private window.
5. Some sites (chrome.google.com, new-tab pages) block extension content scripts by design. Try a regular webpage.

Yes, with Pro. Under Settings → Shortcuts, each custom prompt can be bound to its own key combo — skip the palette entirely for your most-used transformations.

Usually a provider problem. Open the popup — does the badge say On-device, Cloud, or No provider?

• On-device, still fails: Nano isn't ready. See the Nano setup section.
• Cloud, fails: Open Intelligence, click Test on your provider. You'll get a specific error (bad key, rate limit, network).
• No provider: Enable Nano or add an API key.

Some surfaces are read-only — PDFs, some iframes, and sites with strict DOM sandboxing. In those cases the result is copied to your clipboard and shown in a panel; paste it manually with Ctrl+V.

On PDFs specifically, Chrome will ask for clipboard permission the first time — click Allow.

Known edge cases:

• Google Docs: uses a canvas editor — selection works, but replacement happens via clipboard paste.
• LinkedIn / Quill editors: the editor resets selection on focus; replacement may need a manual paste.
• Gmail subject line: works with replacement.
• Stack Overflow / textareas: full support.
• chrome://, Chrome Web Store pages: blocked by Chrome, can't inject.

If a site you care about misbehaves, email support with the URL.

The API key is wrong, revoked, or missing credits. Open Settings → AI Providers, re-paste the key (check for trailing whitespace), and click Test.

You're hitting the provider's rate limit. Wait a minute and retry, or switch to a different provider for heavy bursts.

Try reloading the tab you want to use it on. The content script injects at document-idle — tabs opened before install won't have it until reload. If the popup won't open at all, disable and re-enable ClipAction at chrome://extensions.

Open chrome://extensions → ClipAction → the version number is under the name. Email clipactionsupport@protonmail.com with the version, Chrome version (chrome://version), and steps to reproduce. Attach the browser console output (F12 → Console) if you can.

A custom prompt is a named instruction with an optional keyboard shortcut. Open Settings → Custom Prompts, write an instruction like "Rewrite this in my voice — dry, confident, short sentences: {{selection}}", pick a model, save. It shows up in the palette and runs on any selection. See the Action Builder.

A site profile scopes which actions appear in the palette based on the domain you're on. GitHub gets your PR-review and issue-summarize prompts; Gmail gets tone-softener and reply-drafter; neither sees the other's clutter. Match by exact domain, subdomain, or regex (e.g. ^https://(app|www)\.slack\.com). See the Site Profiles.

Chains run two or more actions in sequence, feeding the output of one into the input of the next. Useful for things like "translate to English → summarize in 3 bullets" or "extract key points → draft reply". Configure them under Settings → Chains.

Four built-in placeholders:

• {{selection}} — the highlighted text (required in every prompt)
• {{language}} — target language, when the action asks for one
• {{tone}} — tone directive (professional, casual, technical…)
• {{length}} — length hint (short, medium, long)

Anything else in the prompt is just plain text.

Not yet — on the roadmap. For now they're stored per-browser. If you need to move them between machines, email support and we'll help.

Locally, in Chrome's chrome.storage.local. That includes your API keys, custom actions, site profiles, and preferences. Nothing is synced to our servers.

To wipe all settings, keys, and prompts, right-click the ClipAction icon in the toolbar and Inspect popup. In the DevTools console that opens, run:

chrome.storage.local.clear().then(() => location.reload())

To re-trigger only the first-run onboarding (keep your keys), run:

chrome.storage.local.remove(['onboardingComplete','onboardingSkipped']).then(()=>location.reload())

Right-click the ClipAction icon in the toolbar → Remove from Chrome. All local data is deleted with the extension.

Email clipactionsupport@protonmail.com. Include your Chrome version, extension version, and what you were trying to do.