Context Copilot

Your API key. Your data. AI that works for you.

Context Copilot — BYO-Key AI on any webpage Bring your own Anthropic (Claude) or OpenAI (GPT) API key. Context Copilot stores it locally in your browser and routes requests directly to the provider. No subscription. No account. No backend server. The AI works for you, not a third party. ───────────────────────────────────────────── HOW IT WORKS ───────────────────────────────────────────── Click the Context Copilot icon to open the side panel on any tab. The extension reads the current page's text automatically — ask questions, run skills, or start a conversation without copying and pasting anything. Right-click any highlighted text to instantly apply a skill from the context menu. Or open the side panel, tap a skill chip, and hit send. Page content is available as context automatically. Everything happens in your browser. Your keystrokes, your page content, and your API key never touch any server except the AI provider you choose. ───────────────────────────────────────────── 10 BUILT-IN SKILLS ───────────────────────────────────────────── Skills are pre-built prompts that shape how the AI responds. Tap a chip to activate one, or right-click highlighted text to apply it instantly. SUMMARISE Condenses long-form content into the key points. Works on articles, research papers, product docs, meeting transcripts, emails, and anything else on the page. Example: Open a dense industry report, tap Summarise, and get a tight paragraph covering the core findings — without reading the whole thing. REWRITE Rewrites selected text in any tone. Professional, casual, direct, confident, empathetic — specify what you need. Useful for refining emails before you send them, tightening website copy, or converting rough notes into polished paragraphs. Example: Paste a stiff client email, type "casual and direct", and get a version that sounds like a person wrote it. EXPLAIN Breaks down complex or technical content into plain language. Works on legal text, scientific abstracts, developer documentation, financial reports, and any jargon-heavy material. Example: Select a dense paragraph from a terms-of-service document, right-click, hit Explain, and get a plain-English version of what you are actually agreeing to. TRANSLATE Translates selected text or full-page content into any language. Supports major world languages including Spanish, French, German, Japanese, Chinese (Simplified and Traditional), Portuguese, Italian, Korean, Arabic, and more. Example: Land on a product page in another language, tap Translate, specify your language, and read the full content without leaving the site. EXTRACT ACTIONS Pulls action items, next steps, decisions, and commitments from any content. Turns messy meeting notes into a clean numbered task list. Works on emails, threads, project briefs, call transcripts, and documents. Example: Paste in notes from a client call, hit Extract Actions, and get a numbered list of who is doing what and by when — ready to drop into your project tool. CHAT Open conversation with the current page as automatic context. Ask follow-up questions, request clarifications, or explore the content without switching tabs. Maintains conversation history within your session. Example: Reading a long product changelog? Ask "what changed in the authentication flow?" and get a direct answer without hunting through the page yourself. SPREADSHEET ANALYSIS Analyses tabular data, generates formulas, interprets trends, and structures information into rows and columns. Works on any page containing tables, numbers, or structured data. Example: Open a competitor pricing page, tap Spreadsheet Analysis, and get a structured comparison table with a plain-language summary of where they sit on value versus price. ALIGNED IDEATION Structured brainstorming that organises ideas across three tiers: Quick Wins (executable this week), Momentum Builders (medium-term plays), and Compound Plays (long-horizon bets). Useful for strategy, product, and content work. Example: Reading about a market shift? Run Aligned Ideation on the article and get a prioritised list of moves you could make at three different time horizons. DEBUG Identifies bugs, explains the root cause, and provides a targeted fix. Works on code visible on any page — Stack Overflow questions, GitHub diffs, documentation, or code you paste in directly. Output is structured: root cause, why it fails, corrected code, adjacent issues spotted, and how to verify the fix. Example: Select a broken useEffect on Stack Overflow, right-click, tap Debug this code, and get a clear diagnosis with the corrected version and an explanation of why the original failed. PRO-PROMPT An optional meta-skill. When enabled, your input is refined by a lightweight model before it reaches the main model. You type naturally — Pro-Prompt adds structure, constraints, and specificity before sending. The result is a sharper, more targeted response without you needing to learn prompt engineering. The comparison between your original input and the improved version is shown alongside the response so you can see exactly what changed. Example: Type "summarise this for my boss" — Pro-Prompt converts it to a structured instruction with format, length, and audience constraints before it hits the main model. ───────────────────────────────────────────── CUSTOM SKILLS ───────────────────────────────────────────── The built-in skills cover common workflows, but your work may have specific recurring needs. Build your own skills in Settings. Give each one a name, an icon, and a system prompt. Custom skills appear as chips in the skill bar and in the right-click context menu alongside the built-in ones. No coding required. If you can describe what you want the AI to do in plain language, you can build a skill. Some examples of what people build: - A legal researcher's "Find Precedents" skill that extracts case references and cites from any article - A copywriter's "Brand Voice" skill that rewrites any text in a specific client's tone and style - A developer's "Code Review" skill tuned to their team's conventions and style guide - A founder's "Competitor Intel" skill that extracts pricing, positioning, and feature signals from any competitor page ───────────────────────────────────────────── WHY BYO-KEY MATTERS ───────────────────────────────────────────── Most AI browser extensions work one of two ways: they charge a subscription and proxy your requests through their own servers, or they offer a free tier and monetise your data. In both cases, a third party is in the middle of every request you make. Context Copilot does not work that way. You bring your own API key from Anthropic or OpenAI. Your key is stored in chrome.storage.local on your device. When you make a request, your browser connects directly to the API provider — there is no Context Copilot server in the path. The developer cannot see your key, your prompts, or your page content, because none of it ever reaches any server the developer operates. The cost is also different. Instead of a fixed monthly subscription, you pay your API provider only for what you use. Typical usage costs fractions of a cent per request. A few dozen requests per day across a month costs roughly the price of a coffee. Heavy usage might cost a few dollars. You are not locked into a subscription, and the extension does not stop working if you forget to renew. ───────────────────────────────────────────── PRIVACY ───────────────────────────────────────────── API keys stored in chrome.storage.local only — transmitted to no server except your chosen AI provider, and only at the moment you make a request. Page content read locally in your browser — sent to your AI provider as part of your prompt, never stored or logged by the developer. Chat history stored in chrome.storage.local on your device — not transmitted to any server operated by the developer. Zero analytics. Zero telemetry. Zero crash reporting. Zero usage tracking of any kind. No account required. No sign-up. No email. No profile. No backend server — there is nothing to breach because there is nothing stored. Full privacy policy is linked in the store listing. ───────────────────────────────────────────── WHO IS THIS FOR ───────────────────────────────────────────── Context Copilot is for people who work with web content every day and are tired of switching tabs to paste things into a separate AI window. RESEARCHERS AND ANALYSTS Summarise papers and reports without reading them end to end. Extract data points from tables. Ask follow-up questions about a source. Translate foreign-language material. Build a picture of a topic faster without losing your train of thought. WRITERS AND EDITORS Rewrite copy in any tone directly on the page you are referencing. Translate your work for international audiences. Summarise source material before drafting. Run a custom brand voice skill on any text. DEVELOPERS Debug code on Stack Overflow, GitHub, or any documentation page without switching tools. Explain unfamiliar technical concepts in plain language. Extract action items from issue threads and changelogs. FOUNDERS AND OPERATORS Extract next steps from meeting notes and email threads the moment you read them. Run Aligned Ideation on market research, competitor pages, or strategy documents. Summarise long reports into briefing notes. STUDENTS AND ACADEMICS Explain complex academic content in plain language. Translate international sources. Summarise long papers into structured notes. Ask questions about a reading without leaving the page. ───────────────────────────────────────────── VOICE INPUT ───────────────────────────────────────────── Tap the microphone button to speak your request instead of typing. Uses Chrome's built-in Web Speech API — no third-party voice service, no audio transmitted to any server operated by the developer. Language configurable in Settings. Supported languages: English (AU, US, UK), Spanish, French, German, Japanese, Chinese (Mandarin), Portuguese (BR). ───────────────────────────────────────────── USAGE TRACKING ───────────────────────────────────────────── A live token count and estimated cost in USD is shown in the panel header while you work. Full breakdown by model is available in Settings, including a per-model split so you can see exactly what each provider is costing you. Reset any time. ───────────────────────────────────────────── SETUP — 60 SECONDS ───────────────────────────────────────────── 1. Install Context Copilot from the Chrome Web Store 2. Click the puzzle piece icon in Chrome's toolbar and pin Context Copilot 3. Open any webpage and click the Context Copilot icon to open the side panel 4. Go to Settings and paste your Anthropic or OpenAI API key 5. Tap any skill chip or start typing in the chat input To get an API key: Anthropic — console.anthropic.com. Create an account, add a payment method, and generate a key under API Keys. OpenAI — platform.openai.com. Same process under API Keys. Both are pay-as-you-go with no minimum spend. You only pay when you use the extension. ───────────────────────────────────────────── FREQUENTLY ASKED QUESTIONS ───────────────────────────────────────────── Does it work on every website? It works on nearly all standard websites. It cannot run on Chrome internal pages (chrome://) or the Chrome Web Store itself — this is a Chrome security restriction that applies to all extensions, not specific to Context Copilot. Which AI models does it use? For Anthropic: Claude Haiku for fast, low-cost tasks and Claude Sonnet for more complex ones. For OpenAI: GPT-4o mini and GPT-4o. The extension selects the appropriate model based on the skill. You can set a default provider preference in Settings. Do I need both an Anthropic and an OpenAI key? No. One key is enough. The extension auto-detects which provider to use based on what key you have saved. Add both if you want access to models from either provider. Is my data safe if I use it on sensitive pages? Your page content is sent to your chosen AI provider as part of your prompt — the same as if you copied it into Claude.ai or ChatGPT manually. It is not stored by the developer. Review Anthropic's or OpenAI's privacy policy for their data handling terms before using sensitive content. What does Pro-Prompt cost? Pro-Prompt makes one additional API call before your main request, using a fast, low-cost model. The additional cost is typically less than 0.001 USD per use. It is included in your usage breakdown in Settings. Can I use my own system prompts? Yes. Build a Custom Skill in Settings and write any system prompt you like. It will appear as a chip alongside the built-in skills and works identically. ───────────────────────────────────────────── REQUIREMENTS ───────────────────────────────────────────── Chrome desktop, version 116 or later (required for side panel support). An Anthropic API key, an OpenAI API key, or both. The extension uses the following permissions: - storage: to save your API key, settings, skills, chat history, and usage stats locally - activeTab: to read the current tab for content script safety checks - scripting: to inject the content script that reads page text on demand - tabs: to look up the active tab ID for content script messaging - contextMenus: to add skill options to the right-click menu - sidePanel: to power the side panel interface - all_urls host permission: so the content script can read page text on any site you visit. Your key is only ever sent to api.anthropic.com or api.openai.com — nowhere else. ───────────────────────────────────────────── Built and maintained by Crow_Code. Open code. No subscription. AI that works for you.