Skip to content

Open-source AI browser agent for Chrome and Firefox (monorepo) 🧠

49.7 KB
Raw Blame
{
  "meta.title": "WebBrain β€” Free Open-Source AI Browser Agent for Chrome, Firefox & Edge",
  "meta.description": "WebBrain is a free, open-source alternative to Claude's browser plugin. AI browser agent for Chrome, Firefox, and Edge β€” read pages, extract data, automate tasks. Works with llama.cpp, OpenAI, Claude, OpenRouter. Multilingual UI, token-conscious screenshots. GPL-3.0-or-later licensed.",
  "meta.keywords": "AI browser agent, browser automation, open source browser plugin, Claude browser plugin alternative, OpenClaw alternative, web automation, LLM browser extension, chrome extension AI, firefox AI addon, llama.cpp browser, GPT browser agent, browser AI assistant, web scraping AI, open source AI tool, self-hosted AI browser, multilingual browser AI, token-efficient AI extension",
  "meta.og_title": "WebBrain β€” Free Open-Source AI Browser Agent",
  "meta.og_description": "The open-source alternative to Claude's browser plugin. AI-powered browser extension for Chrome & Firefox. Read, understand, and automate any web page. Multi-provider LLM support. Multilingual UI. Free forever.",
  "meta.twitter_title": "WebBrain β€” Open-Source AI Browser Agent",
  "meta.twitter_description": "Free, open-source alternative to Claude browser plugin. AI agent for Chrome & Firefox. Works with local or cloud LLMs. 5-language UI. Token-conscious by default.",
  "meta.image_alt": "WebBrain icon",
  "meta.software_version": "5.1.0",
  "nav.features": "Features",
  "nav.demo": "Demo",
  "nav.providers": "Providers",
  "nav.compare": "Compare",
  "nav.download": "Download",
  "nav.faq": "FAQ",
  "nav.blog": "Blog",
  "nav.docs": "Docs",
  "nav.security": "Security",
  "banner.top_mcp": "MCP server for Claude Code, Codex & Cursor",
  "nav.language_aria": "Select language",
  "nav.github": "GitHub",
  "nav.github_stars_aria": "Star WebBrain on GitHub",
  "nav.theme_toggle_aria": "Toggle light / dark theme",
  "nav.browser_extension": "Browser extension",
  "nav.for_websites": "For websites",
  "nav.research": "Research",
  "home.nav_aria": "Main navigation",
  "home.meta.title": "WebBrain β€” AI agent for your browser and your website",
  "home.meta.description": "WebBrain is a free AI agent that reads, explains, and acts on web pages. Add the open-source extension to Chrome, Edge, or Firefox, or add WebBrain to your website for your visitors.",
  "home.meta.og_title": "WebBrain β€” An AI agent on every page you visit or build",
  "home.kicker": "Free for browsers and websites",
  "home.title_line1": "An AI agent on every page",
  "home.title_line2": "you visit or build.",
  "home.lede": "WebBrain reads, explains, and acts on web pages. Add it to your browser to get help on any site, or to your website so your visitors get the same help.",
  "home.kicker_browser": "Free and open source",
  "home.title_line1_browser": "An AI agent on every page",
  "home.title_line2_browser": "you visit.",
  "home.lede_browser": "WebBrain reads, explains, and acts on web pages. Add it to Chrome, Edge, or Firefox and get help on any site, with any model, including local ones.",
  "home.browser.title": "In your browser",
  "home.browser.desc": "An open-source AI side panel for Chrome, Edge, and Firefox. Chat with any page, pull out data, and automate repetitive work β€” with any model, including local ones.",
  "home.browser.badge": "Open Source",
  "home.browser.watch": "Watch it in action",
  "home.browser.features": "Features",
  "home.websites.title": "On your website",
  "home.websites.badge": "Free",
  "home.websites.desc": "One script tag gives every visitor an AI guide that explains your pages, answers questions, and fills in forms when you allow it.",
  "home.websites.cta": "Add WebBrain to your website",
  "home.websites.demo": "Demo",
  "home.websites.info": "Info",
  "home.websites.demo_error": "The demo couldn't load. Try again in a moment.",
  "home.video.close_aria": "Close the video",
  "hero.h1_html": "The Open-Source AI <span class=\"gradient\">Browser Agent</span>",
  "hero.description": "Free, open-source alternative to Claude in Chrome. Use any model, including local ones, so your pages never leave your machine.",
  "hero.cta.install": "Add to Chrome β€” free",
  "hero.cta.github": "View on GitHub",
  "hero.proof.chrome_featured": "Featured on Chrome Web Store",
  "hero.proof.product_hunt": "Product Hunt",
  "hero.proof.users": "Thousands of users across browsers",
  "hero.trust.b1": "Read-only Ask Mode by default",
  "hero.trust.b2": "Asks before consequential actions",
  "hero.trust.b3": "Open-source & auditable",
  "hero.trust.b4": "Choose local or cloud models",
  "hero.mockup.url": "https://example.com/products",
  "hero.mockup.page_title": "Product Catalog",
  "hero.mockup.user_prompt": "Extract all product names and prices from this page",
  "hero.mockup.step_reading": "Reading page",
  "hero.mockup.step_extracting": "Extracting data",
  "hero.mockup.result_html": "Found <strong>24 products</strong>. Here are the results:",
  "hero.mockup.result_more": "...and 21 more",
  "hero.mockup.input_placeholder": "Ask anything about this page...",
  "hero.mockup.scene_nav_aria": "Switch demo scenario",
  "hero.mockup.scene_prev_aria": "Previous demo scenario",
  "hero.mockup.scene_next_aria": "Next demo scenario",
  "hero.mockup.scene_label.products": "Product extraction demo",
  "hero.mockup.tab.followup": "Summarize",
  "hero.mockup.tab.form": "Fill a form",
  "hero.mockup.tab.spacing": "Web dev",
  "hero.mockup.tab.video": "Save a video",
  "hero.mockup.tab.products": "Automate",
  "hero.mockup.scene_label.website": "Clone-this-site demo",
  "hero.mockup.scene_label.video": "Video download demo",
  "hero.mockup.scene_label.followup": "Follow-up question demo",
  "hero.mockup.v2.url": "https://acmestudio.io",
  "hero.mockup.v2.brand": "Acme Studio",
  "hero.mockup.v2.headline_html": "Build <em>beautiful</em> things",
  "hero.mockup.v2.user_prompt": "Make a site like this for my brand \"Lumen Coffee\"",
  "hero.mockup.v2.step_layout": "Analyzing layout & sections",
  "hero.mockup.v2.step_tokens": "Extracting fonts & colors",
  "hero.mockup.v2.result_html": "Generated <strong>Lumen Coffee</strong> with matching layout:",
  "hero.mockup.v2.result_palette": "Hero Β· Features Β· Pricing",
  "hero.mockup.v2.result_link": "Open in editor β†’",
  "hero.mockup.v3.url": "https://youtube.com/watch?v=n0m4dB0druM",
  "hero.mockup.v3.username": "@nomad",
  "hero.mockup.v3.caption": "Sunset timelapse over Bodrum",
  "hero.mockup.v3.user_prompt": "Download this video",
  "hero.mockup.v3.step_locate": "Locating video stream",
  "hero.mockup.v3.step_fetch": "Fetching MP4",
  "hero.mockup.v3.result_html": "Saved to <strong>Downloads/nomad-bodrum.mp4</strong>",
  "hero.mockup.v3.result_meta": "1080p Β· 24s Β· 4.2 MB",
  "hero.mockup.v3.download_complete": "Download complete",
  "hero.mockup.v4.url": "https://magazine.dev/local-ai-2026",
  "hero.mockup.v4.page_title": "The Future of Local AI",
  "hero.mockup.v4.byline": "By Jane Smith Β· 8 min read",
  "hero.mockup.v4.user_prompt_1": "Summarize this article",
  "hero.mockup.v4.assistant_1_html": "Local LLMs now match cloud quality on many tasks β€” driven by better quantization, faster consumer GPUs, and a privacy-first shift.",
  "hero.mockup.v4.user_prompt_2": "What does the author predict for 2026?",
  "hero.mockup.v4.step_recall": "Recalling article context",
  "hero.mockup.v4.assistant_2_html": "By <strong>2026</strong>, the author expects 70%+ of consumer AI workloads to run on-device.",
  "hero.mockup.scene_label.form": "Form filling demo",
  "hero.mockup.scene_label.spacing": "CSS spacing demo",
  "hero.mockup.products.subtitle": "248 items Β· Sorted by bestselling",
  "hero.mockup.form.url": "https://tax.example.gov/filing/quarterly-return",
  "hero.mockup.form.title": "Quarterly Tax Return β€” Form QT-4",
  "hero.mockup.form.subtitle": "Department of Revenue Β· All fields required",
  "hero.mockup.form.mem": "from memory",
  "hero.mockup.form.f1a": "Legal business name",
  "hero.mockup.form.f1b": "EIN / Tax ID",
  "hero.mockup.form.f2": "Filing period",
  "hero.mockup.form.f5a": "Gross revenue",
  "hero.mockup.form.f5b": "Deductible expenses",
  "hero.mockup.form.f7": "Total payroll wages",
  "hero.mockup.form.f9a": "Estimated payments made",
  "hero.mockup.form.f9b": "Prior-period credit",
  "hero.mockup.form.f11": "Net tax due",
  "hero.mockup.form.submit": "Submit return",
  "hero.mockup.form.user_prompt": "Fill out this form",
  "hero.mockup.form.step_memory_html": "Recalled name, EIN &amp; payroll <strong>from memory</strong>",
  "hero.mockup.form.step_fields": "Found 9 required fields",
  "hero.mockup.form.step_totals": "Computing totals from figures",
  "hero.mockup.form.result_html": "Filled <strong>9 fields</strong>. Stopping before submit β€” review and send when ready.",
  "hero.mockup.dev.url": "http://localhost:3000/preview",
  "hero.mockup.dev.user_prompt": "How could I make these blocks farther away from each other, with more spacing?",
  "hero.mockup.dev.step_inspect": "Inspecting the grid in DevTools",
  "hero.mockup.dev.result_html": "Bump the grid <strong>gap</strong> and give the section some margin.",
  "demo.label": "Demo",
  "demo.title": "Watch WebBrain in action",
  "demo.subtitle": "See how WebBrain reads pages, extracts data, and automates browser tasks.",
  "security.label": "Security",
  "security.title": "How WebBrain keeps you in control",
  "security.subtitle": "The browser is where real work happens β€” and an adversarial surface. WebBrain starts in read-only Ask Mode, asks before it acts, defends against hidden prompt injections, and handles sensitive fields with stricter rules.",
  "security.video.open_aria": "Watch the security video",
  "security.video.close_aria": "Close the security video",
  "features.label": "Features",
  "features.show_more": "Show more features",
  "features.show_less": "Show fewer",
  "features.title": "Everything you need in a browser AI",
  "features.subtitle": "A full-featured AI agent that lives in your browser sidebar and understands any web page.",
  "apocalypse.label": "Apocalypse Mode",
  "apocalypse.title": "WebBrain, ready when the internet isn’t.",
  "apocalypse.heading": "Offline knowledge, under your control",
  "apocalypse.description": "Set up the four essentials for private local chat: a Text Model, a Vision Model, one offline Wikipedia edition, and the Emergency Box.",
  "apocalypse.video.cta": "Watch our video",
  "apocalypse.video.close_aria": "Close the Apocalypse Mode video",
  "features.page_understanding.title": "Page Understanding",
  "features.page_understanding.desc": "Reads and comprehends any web page β€” articles, docs, dashboards, forms. Ask questions and get instant answers from the current page content.",
  "features.agent.title": "Full Browser Agent",
  "features.agent.desc": "Click, type, scroll, navigate, and interact with pages on your behalf. Automate repetitive tasks with natural language instructions.",
  "features.agent.note": "Asks before consequential actions; read-only Ask Mode is the default.",
  "features.extraction.title": "Data Extraction",
  "features.extraction.desc": "Extract structured data from any page β€” tables, lists, links, forms. Export product catalogs, search results, or any page content. Works with PDFs.",
  "features.multi_provider.title": "Multi-Provider LLM",
  "features.multi_provider.desc": "Works with local llama.cpp, OpenAI, Claude, and OpenRouter. Use your preferred model β€” or run completely offline with local AI.",
  "features.privacy.title": "Privacy First",
  "features.privacy.desc": "Requests to local models stay on your device. Cloud providers receive the content you send. Compass’s β€œHelp Improve WebBrain” is on by default and may share eligible interactions for improvement and training. Turn it off in Settings. Fully open source.",
  "features.smart_context.title": "Smart Context",
  "features.smart_context.desc": "Automatic context management prevents token overflow. Intelligently trims conversation history and limits tool output for smooth, uninterrupted sessions.",
  "features.vision.title": "Dedicated Vision Model",
  "features.vision.desc": "Pair a fast text-only model for planning with a separate vision-capable model for reading screenshots. Cheaper and faster than using a single big multimodal model for everything.",
  "features.profile.title": "Profile Auto-fill",
  "features.profile.desc": "Optional plaintext bio β€” name, work email, company, a throwaway password β€” lets the agent breeze through low-stakes signup forms without asking every time. Off by default, all stored locally.",
  "features.profile.note": "Off by default. Stored locally in your browser, never sent to WebBrain.",
  "features.cookies.title": "Cookie & Paywall Aware",
  "features.cookies.desc": "Dismisses consent banners (OneTrust, Cookiebot, Didomi, Quantcast) before reasoning about the page. Detects paywalls and tells you honestly instead of fabricating article content or trying to bypass them.",
  "features.multilingual.title": "Multilingual UI",
  "features.multilingual.desc": "The plugin ships in English, EspaΓ±ol, FranΓ§ais, TΓΌrkΓ§e and δΈ­ζ–‡. Auto-detects your browser language on first use; switch anytime from the globe icon in the side panel. The marketing site is localized to match.",
  "features.token_conscious.title": "Token-Conscious",
  "features.token_conscious.desc": "Screenshots are resized and iteratively JPEG-compressed before they leave your machine, keeping image tokens small. Smart context trimming and tool-output caps keep cloud bills predictable β€” no surprise spend on long sessions.",
  "providers.label": "LLM Providers",
  "providers.title": "Bring your own AI",
  "providers.subtitle": "Connect to any OpenAI-compatible API or run local models with Ollama, LM Studio, or llama.cpp. Keep your pages, prompts, and browsing data entirely on your machine.",
  "featured_home_promo.label": "Ollama launch handoff",
  "featured_home_promo.title": "Choose an Ollama model, then hand it to WebBrain",
  "featured_home_promo.body": "A new launch path can configure WebBrain for a local Ollama model from one terminal command. It is available today from our Ollama branch while we work toward upstream integration.",
  "featured_home_promo.cta": "Read the announcement",
  "featured_home_promo.image_alt": "WebBrain logo, a heart, and the Ollama logo for the launch handoff announcement",
  "modes.label": "Interaction Modes",
  "modes.title": "Ask or Act",
  "modes.subtitle": "Two modes for different needs. Read-only by default, full agent power when you need it.",
  "modes.ask.title": "Ask Mode",
  "modes.ask.desc": "Read-only. Ask questions about the current page, extract information, summarize content. Safe and non-intrusive β€” nothing gets modified.",
  "modes.act.title": "Act Mode",
  "modes.act.desc": "Full agent. Click buttons, fill forms, navigate between pages, run scripts. Automate complex multi-step browser workflows with a single instruction.",
  "download.label": "Get Started",
  "download.title": "Install WebBrain",
  "download.subtitle": "Available for Chrome, Edge, and Firefox. Free, open-source, no account required.",
  "download.chrome.name": "Chrome & Chromium",
  "download.chrome.desc_html": "Manifest V3 Β· Chrome 116+ Β· Also works with Brave, Opera, Vivaldi, and other Chromium-compatible browsers.",
  "download.chrome.install": "Download from Chrome Web Store",
  "download.chrome.source": "View Chrome source",
  "download.edge.name": "Microsoft Edge",
  "download.edge.desc_html": "Manifest V3 Β· Windows 10/11 Β· Requires Microsoft Edge on Windows.",
  "download.edge.install": "Download from Edge Add-ons",
  "download.edge.source": "View Chromium source",
  "download.firefox.name": "Firefox",
  "download.firefox.desc_html": "Manifest V2 Β· Firefox 109+",
  "download.firefox.install": "Download from Firefox Add‑ons",
  "download.firefox.source": "View Firefox source",
  "cloud_strip.badge": "New",
  "cloud_strip.label": "For AI agents",
  "cloud_strip.description": "A private, visible browser for Codex and other AI agents.",
  "cloud_strip.cta": "Explore Cloud",
  "mcp.label": "New Β· MCP server",
  "mcp.title": "Your coding agent, inside the browser you already use.",
  "mcp.subtitle": "Connect Claude Code, Codex, Cursor, OpenCode, or any stdio MCP client to your signed-in Chromium tabs. WebBrain keeps the task-level permission gate in the browser β€” and now returns structured JSON too.",
  "mcp.docs_cta": "Open the complete MCP guide",
  "mcp.step1_title": "Ask your agent to set it up",
  "mcp.step1_body": "Paste this into Claude Code, Codex, Cursor, or any local MCP client. It reads the guide and adds WebBrain for you.",
  "mcp.prompt": "Read https://webbrain.one/docs/mcp/ and set up the WebBrain MCP server for me.",
  "mcp.copy": "Copy",
  "mcp.copied": "Copied",
  "mcp.step2_title": "Then ask for browser work",
  "mcp.step2_body": "Your agent hands the job to WebBrain in the browser you are already signed into. WebBrain still asks before consequential actions.",
  "mcp.example": "Use WebBrain to summarize the CI dashboard open in my browser.",
  "mcp.requirements": "Needs Chrome, Edge, Brave, Opera, or Vivaldi with WebBrain installed, and Node.js 20+ on the same computer.",
  "compare.label": "Why WebBrain?",
  "compare.title": "How does WebBrain compare?",
  "compare.subtitle": "WebBrain sits at the intersection of browser-native AI plugins and full agent frameworks. Here's how it stacks up.",
  "compare.speed_video.title": "Local WebBrain beats Claude in Chrome, side by side",
  "compare.speed_video.description": "Same task, same browser. WebBrain runs Gemma 4 31B on-device and finishes first.",
  "compare.speed_video.cta": "Watch the comparison",
  "compare.speed_video.close_aria": "Close the comparison video",
  "compare.vs_plugins.heading": "vs. Browser AI Plugins",
  "compare.vs_frameworks.heading_html": "vs. AI Agent Frameworks <span style=\"font-weight:400; font-size:13px;\">(different category)</span>",
  "compare.col_feature": "Feature",
  "compare.col_webbrain": "WebBrain",
  "compare.col_claude": "Claude in Chrome",
  "compare.col_aspect": "Aspect",
  "compare.col_frameworks": "OpenClaw / Browser-Use / etc.",
  "compare.row.open_source": "Open Source",
  "compare.cell.license": "GPL-3.0-or-later License",
  "compare.cell.proprietary": "Proprietary",
  "compare.row.price": "Price",
  "compare.cell.free_forever": "Free forever",
  "compare.cell.claude_pro": "Requires Claude Pro ($20/mo)",
  "compare.row.local_llm": "Local LLM support",
  "compare.cell.local_llm_yes": "llama.cpp, Ollama",
  "compare.cell.local_llm_no": "No β€” Claude only",
  "compare.row.multi_provider": "Multi-provider",
  "compare.cell.multi_provider_yes": "All OpenAI-compatible endpoints",
  "compare.cell.multi_provider_no": "Claude only",
  "compare.cell.yes": "Yes",
  "compare.cell.no": "No",
  "compare.row.side_panel": "Side panel UI",
  "compare.row.offline": "Fully offline",
  "compare.cell.offline_yes": "Yes (with local LLM)",
  "compare.cell.offline_no": "No β€” cloud required",
  "compare.row.self_host": "Self-hostable",
  "compare.row.what_is_it": "What is it?",
  "compare.cell.extension_end_user": "Browser extension (end-user tool)",
  "compare.cell.framework_sdk": "Agent framework / SDK (developer tool)",
  "compare.row.target_user": "Target user",
  "compare.cell.anyone": "Anyone β€” no coding needed",
  "compare.cell.developers": "Developers building automations",
  "compare.row.install": "Installation",
  "compare.cell.one_click": "One-click browser install",
  "compare.cell.python_docker": "Python/Docker setup required",
  "compare.row.ui": "UI",
  "compare.cell.builtin_panel": "Built-in side panel chat",
  "compare.cell.no_ui": "No UI β€” code or API only",
  "compare.row.browser_control": "Browser control",
  "compare.cell.content_script_lightweight": "Content script (lightweight)",
  "compare.cell.cdp_full": "CDP / Playwright (full control)",
  "compare.row.multi_tab": "Multi-tab workflows",
  "compare.cell.per_tab_chats": "Per-tab conversations",
  "compare.cell.programmable_multitab": "Programmable multi-tab orchestration",
  "compare.row.headless": "Headless mode",
  "compare.cell.headless_no": "No β€” runs in your browser",
  "compare.cell.headless_yes": "Yes β€” headless automation",
  "compare.row.extensibility": "Extensibility",
  "compare.cell.custom_providers": "Add custom LLM providers",
  "compare.cell.python_sdk": "Full Python SDK, custom tools",
  "compare.row.best_for": "Best for",
  "compare.cell.daily_ai": "Daily browsing AI assistant",
  "compare.cell.scraping_testing": "Automated scraping / testing pipelines",
  "compare.disclaimer": "WebBrain is a browser extension for end users who want an AI assistant while they browse. Agent frameworks like OpenClaw are developer tools for building automated browser pipelines. Different tools for different jobs β€” and you can use both.",
  "faq.label": "FAQ",
  "faq.title": "Frequently Asked Questions",
  "faq.page.skip_to_main": "Skip to main content",
  "faq.page.lede": "Quick answers about setup, models, billing, safety, and how WebBrain behaves in your browser.",
  "faq.page.search": "Search questions and answers…",
  "faq.page.no_results": "No answers match that search.",
  "faq.page.topics": "Topics",
  "faq.category.about": "About WebBrain",
  "faq.category.cloud": "Cloud and billing",
  "faq.category.models": "Models and local setup",
  "faq.category.browsers": "Browsers and everyday use",
  "faq.category.safety": "Safety and privacy",
  "faq.alt_claude.q": "Is WebBrain a free alternative to Claude's browser plugin?",
  "faq.alt_claude.a_html": "Yes. WebBrain provides similar AI browser agent capabilities β€” reading pages, extracting data, clicking buttons, filling forms, and automating multi-step workflows. Unlike Claude's proprietary browser plugin which requires a Claude Pro subscription and only works with Anthropic's models, WebBrain is completely free, open-source (GPL-3.0-or-later license), and supports multiple LLM providers including local models that run entirely on your machine.",
  "faq.cloud_subscription.q": "How much is a WebBrain Compass subscription?",
  "faq.cloud_subscription.a_html": "<p>WebBrain Compass is currently <strong>$5/month per device profile</strong>, much cheaper than Claude Pro for the time being. Usage is currently unlimited under a fair usage policy: normal personal use is fine, but abusive, automated, resale, or unusually high-volume use is not allowed.</p><p>The subscription is linked to the device identifier generated by the extension for that browser + OS GUID, not to a user account. If that identifier is lost because you reset extension storage, delete the browser profile, reinstall the browser, change browser/OS, or lose the device, we cannot retrieve or transfer it and we do not issue refunds for that loss.</p>",
  "faq.manage_subscription.q": "How do I manage or cancel my subscription?",
  "faq.manage_subscription.a_html": "<p>Manage billing at <a href=\"https://api.webbrain.one/account\" target=\"_blank\" rel=\"noopener\">api.webbrain.one/account</a> and use your email to log in.</p>",
  "faq.subscription_unrecognized.q": "I already paid for WebBrain Compass, but the extension is asking me to pay again. Do I need to pay twice?",
  "faq.subscription_unrecognized.a_html": "<p>No β€” do not pay again. Your Compass subscription is linked to the browser/device identifier generated by the extension, so if that identifier changed (new browser or device, reinstalled browser, deleted browser profile, or reset extension storage), the extension no longer recognizes your existing subscription and asks for payment.</p><p>Reattach your existing subscription to your current browser through the self-service portal at <a href=\"https://api.webbrain.one/portal\" target=\"_blank\" rel=\"noopener\">api.webbrain.one/portal</a> and follow the instructions there to restore access.</p><p>For receipts and other billing changes, manage billing at <a href=\"https://api.webbrain.one/account\" target=\"_blank\" rel=\"noopener\">api.webbrain.one/account</a> with your email.</p>",
  "faq.cloud_sync.q": "How does encrypted Cloud Sync work?",
  "faq.cloud_sync.a_html": "<p>Cloud Sync is an optional feature for active WebBrain Compass subscribers. It syncs your WebBrain memories, Profile auto-fill text, and supported provider configuration, including API keys. It does <strong>not</strong> sync chat history, browser history, page screenshots, or legacy OAuth access and refresh tokens.</p><p>Your browser encrypts the sync vault before upload with a password you choose. WebBrain Compass stores only the encrypted vault; it cannot read your memories, profile text, provider settings, or API keys, and it cannot recover the sync password for you.</p><p>The email field is for WebBrain Compass magic-link billing/subscription authentication. It is not an account password. If the UI says change password in Cloud Sync, it means the <strong>sync encryption password</strong>, not a WebBrain account password.</p><p>If you forget the sync password, the old encrypted cloud vault cannot be decrypted. You can reset the cloud vault and upload a new encrypted copy from a device that still has the data locally.</p>",
  "faq.vs_frameworks.q": "How does WebBrain compare to OpenClaw, Browser-Use, and other AI agent frameworks?",
  "faq.vs_frameworks.a_html": "They're different categories of tools. WebBrain is a browser extension β€” you install it in Chrome or Firefox and chat with it in a side panel, no coding required. Frameworks like OpenClaw and Browser-Use are developer SDKs for building automated browser pipelines in Python, typically using headless browsers and CDP. Think of it this way: WebBrain is for daily browsing with an AI assistant; agent frameworks are for building scraping bots and test automation. You can use both β€” they complement each other.",
  "faq.workflow_editor.q": "Can I edit a WebBrain workflow?",
  "faq.workflow_editor.a_html": "<p>Yes. Use the <a href=\"https://www.webbrain.one/workflow-editor\">WebBrain Workflow Editor</a> to inspect, edit, validate, and export a workflow JSON file in your browser.</p>",
  "faq.offline.q": "Can I use WebBrain completely offline?",
  "faq.offline.a_html": "Yes. WebBrain's default provider is llama.cpp, which runs a local AI model on your computer. No API keys needed, no internet required for the AI, and no data ever leaves your machine. Just download a GGUF model, start llama-server, and you have a fully private AI browser agent. You can also use Ollama with its OpenAI-compatible endpoint.",
  "faq.emergency_box.q": "What is the Emergency Box, and how does offline RAG work?",
  "faq.emergency_box.a_html": "<p>The Emergency Box is a curated collection of ~570 public-domain field-reference documents (medical, survival, education, communication) that WebBrain can download and index locally. Once installed, you can ask factual questions offline and get answers sourced from those documents β€” no internet connection required.</p><p>Under the hood, WebBrain uses SQLite FTS5 for fast keyword search across both the Emergency Box and any installed Wikipedia archives, plus an optional semantic search layer using a locally downloaded multilingual E5 model (~140 MB). Results from both methods are combined and diversified, and every answer cites the exact source passage. Citations open in a local reader that verifies document integrity before displaying β€” they never navigate to a live web page.</p><p>You can also install Wikipedia archives through Apocalypse Mode for a broader offline knowledge base. Source and language filters let you control which collections are searched.</p>",
  "faq.models_supported.q": "Which AI models does WebBrain support?",
  "faq.models_supported.a_html": "WebBrain supports four provider types: llama.cpp (any local GGUF model), OpenAI (GPT-4o, GPT-4, etc.), Claude (Claude Opus, Sonnet, Haiku via native API), and OpenRouter (access to 100+ models from various providers). Any OpenAI-compatible API endpoint works, so you can also use services like Together AI, Groq, Mistral, or any local server with an OpenAI-compatible interface.",
  "faq.recommended_model.date": "April 21, 2026",
  "faq.recommended_model.q": "What's the most recommended model?",
  "faq.recommended_model.a_html": "<p>As of <strong>April 21, 2026</strong>, the top recommendation is <a href=\"https://huggingface.co/Qwen/Qwen3.6-35B-A3B\" target=\"_blank\">Qwen 3.6 35B</a>. Why: in our vision benchmark (<a href=\"https://www.webbrain.one/blog/vision-model-shootout\" target=\"_blank\">vision-model-shootout</a>), it outperformed Gemma 4 on screenshot understanding while remaining practical for local inference.</p><p>For consumer GPUs, RTX 5090 is ideal, and RTX 4090 is often workable with INT4 AutoRound quantization via <a href=\"https://huggingface.co/Intel/Qwen3.6-35B-A3B-int4-AutoRound\" target=\"_blank\">Intel/Qwen3.6-35B-A3B-int4-AutoRound</a>.</p><p>For max speed, we recommend serving on <strong>vLLM</strong>. Sample command:</p><p><code>python -u -m vllm.entrypoints.openai.api_server --model Intel/Qwen3.6-35B-A3B-int4-AutoRound --served-model-name qwen3.6-35b --quantization auto --dtype bfloat16 --max-model-len 65536 --max-num-batched-tokens 32768 --max-num-seqs 4 --host 0.0.0.0 --port 8000 --gpu-memory-utilization 0.92 --enable-prefix-caching --enable-chunked-prefill --limit-mm-per-prompt '{\"image\": 4, \"video\": 1}' --mm-processor-cache-type shm --reasoning-parser qwen3 --enable-auto-tool-choice --tool-call-parser qwen3_coder --trust-remote-code --allowed-origins '[\"*\"]' --speculative-config '{\"method\": \"dflash\", \"model\": \"z-lab/Qwen3.6-35B-A3B-DFlash\", \"num_speculative_tokens\": 15}' --attention-backend flash_attn</code></p><p><em>DFlash speculative decoding is optional.</em></p>",
  "faq.webgpu_apocalypse.q": "Why is WebGPU only available in Apocalypse Mode?",
  "faq.webgpu_apocalypse.a_html": "<p>Because we do not think browser-based WebGPU text inference is good enough yet for most users, and we do not want you to suffer through a poor experience. If you want to use the same LFM2.5 2.6B model locally, run inference on your machine with <strong>llama.cpp</strong> or <strong>Ollama</strong> and point WebBrain to that local endpoint instead of relying on WebGPU.</p><p>The WebGPU vision fallback remains available because its model is much smaller and the fallback is used only occasionally.</p>",
  "faq.cors.q": "I'm getting \"Failed to fetch\" when connecting to a local LLM server (vLLM, Ollama, llama.cpp) on my network",
  "faq.cors.a_html": "<p>If your LLM server is on a different machine on your local network (e.g. <code>http://192.168.1.x:8000</code>), Chrome blocks the request unless the server sends <strong>CORS headers</strong>. The fix depends on your server:</p><p><strong>vLLM:</strong> Start with <code>--allowed-origins '[\"*\"]'</code> (the value must be a JSON list).<br><strong>Ollama:</strong> Set the environment variable <code>OLLAMA_ORIGINS=*</code> before starting.<br><strong>llama.cpp:</strong> CORS is enabled by default β€” no changes needed.</p><p>If your server runs on <code>localhost</code> (same machine as the browser), CORS is usually not required. The issue only affects cross-machine connections on the local network. Make sure your base URL in WebBrain settings ends with <code>/v1</code> (e.g. <code>http://192.168.1.47:8000/v1</code>). Exception: Ollama used from a browser extension may still need <code>OLLAMA_ORIGINS</code>; see the Ollama FAQ below.</p>",
  "faq.ollama_origins.q": "Why does Ollama on localhost return 403 from the WebBrain extension?",
  "faq.ollama_origins.a_html": "<p>Recent Ollama versions can reject requests from browser-extension origins even when Ollama is running on the same machine. The extension sends an <code>Origin</code> such as <code>chrome-extension://...</code> or <code>moz-extension://...</code>, and Ollama may answer 403 unless those origins are allowed.</p><p>Quit any Ollama desktop app that is already holding port <code>11434</code>, then start Ollama with one of these commands:</p><p><code>OLLAMA_ORIGINS=\"*\" ollama serve</code><br><code>OLLAMA_ORIGINS=\"chrome-extension://*,moz-extension://*\" ollama serve</code></p><p>Keep WebBrain's Ollama base URL set to <code>http://localhost:11434/v1</code>. Terminal <code>curl</code> checks can still work without this setting because they do not send a browser-extension origin.</p>",
  "faq.firefox.q": "Does WebBrain work on Firefox?",
  "faq.firefox.a_html": "Yes. WebBrain ships with both a Chrome version (Manifest V3, using the sidePanel API) and a Firefox version (Manifest V2, using sidebar_action). Both versions have identical features. The Firefox version can be loaded as a temporary add-on for development, or published to addons.mozilla.org for permanent installation.",
  "faq.firefox_sidebar_move.q": "Can I move the Firefox sidebar from the left to the right, like Chrome's side panel?",
  "faq.firefox_sidebar_move.a_html": "Yes β€” Firefox's sidebar defaults to the left, but you can flip it. Right-click anywhere in the sidebar header and choose <strong>Move Sidebar to Right</strong> (or use <strong>View β†’ Sidebar β†’ Move Sidebar to Right</strong> from the menu bar). The position persists across restarts. Chrome's sidePanel defaults to the right and isn't user-movable from the panel itself.",
  "faq.vivaldi_dialogs.q": "Are there any limitations when using WebBrain in Vivaldi's Web Panel?",
  "faq.vivaldi_dialogs.a_html": "<p>Yes. Vivaldi Web Panels suppress native JavaScript dialogs such as <code>confirm()</code>, <code>alert()</code>, and <code>prompt()</code>. WebBrain's Ask, Act, and Dev mode switch no longer depends on those dialogs, but a few secondary actions still do.</p><p><strong>Recording:</strong> Do not use <code>/record</code> in Vivaldi for now. If recording fails, Vivaldi can hide the error alert and leave no visible explanation.</p><p><strong>Settings, History, Traces, and Ollama handoff:</strong> These also contain native dialogs, but WebBrain normally opens them as regular tabs, where they work. If you manually place one in a Web Panel, open it in a normal tab before using dialog-dependent actions.</p>",
  "faq.safe.q": "Is WebBrain safe to use? Can it modify web pages?",
  "faq.safe.a_html": "WebBrain has three modes. Ask mode (default) is read-only and cannot modify anything on the page. Act mode enables browser actions such as clicking, typing, and navigating, while Dev mode adds page-debugging tools. Consequential actions require approval by default; if that permission gate is turned off, WebBrain displays a visible warning banner. You can stop the agent at any time with the Stop button. The extension's source code is fully open for audit on GitHub.",
  "faq.disable_approval_questions.q": "It's asking me too many questions. Can I disable that?",
  "faq.disable_approval_questions.a_html": "Those approval prompts are for your safety. Even on sites you trust, malicious actors can mimic LLM instructions or inject content that tries to make the agent do things on your behalf that you would not approve. That is why WebBrain asks before consequential actions. If you still do not want to go through those prompts, type <code>/dangerously-skip-permissions</code> in the side panel, or open the side plugin, click the wheel icon in the header, go to the <strong>Permissions</strong> tab in <code>settings.html</code>, and switch off <strong>Ask before consequential actions</strong>.",
  "faq.scraping.q": "How do I use WebBrain for web scraping and data extraction?",
  "faq.scraping.a_html": "Simply open any web page, open the WebBrain side panel, and ask in natural language: \"Extract all product names and prices from this page\", \"Get all email addresses on this page\", or \"Summarize this article in bullet points\". The AI agent reads the page content, understands the structure, and returns the extracted data. For more complex scraping, switch to Act mode and the agent can navigate between pages, click pagination buttons, and aggregate data across multiple pages.",
  "faq.api_mutations.q": "Does WebBrain call APIs directly, or does it always click through the UI?",
  "faq.api_mutations.a_html": "<p>By default, WebBrain <strong>always goes through the visible UI</strong> for any action that creates, modifies, deletes, sends, submits, posts, or buys anything. It will navigate to the page, fill the form, and click the button β€” exactly the way you would. It refuses to call REST/GraphQL endpoints directly via background <code>fetch()</code> for mutations. This is deliberate: API actions are invisible (you don't see what's being sent), often require separate auth tokens you may not have configured, and have a much larger blast radius than a visible mis-click. UI-first means everything is on screen, in your normal browser session, and stoppable.</p><p>For <strong>reading</strong> data β€” fetching a README, looking up an issue, comparing prices across sites, checking a status page β€” WebBrain freely uses background HTTP requests via the <code>fetch_url</code> and <code>research_url</code> tools. Reading is not the same as acting; it doesn't change anything on a remote service, so the safety concerns don't apply.</p><p>If you specifically want to allow API mutations for a particular task, type <code>/allow-api</code> at the start of your message (optionally followed by a short task description). This per-conversation override lets WebBrain fall back to API endpoints when the UI is genuinely failing or unworkable, while still preferring UI when UI works. A sticky badge stays visible above the input area while the override is active, and it clears when you reset the conversation.</p>",
  "faq.lm_studio.q": "Can I use it in LM Studio too?",
  "faq.lm_studio.a_html": "Yes. WebBrain's read-only network tools β€” <code>fetch_url</code> and <code>research_url</code> β€” also ship as a standalone <a href=\"https://lmstudio.ai\" target=\"_blank\">LM Studio</a> plugin at <a href=\"https://lmstudio.ai/webbrain/web-tools\" target=\"_blank\">webbrain/web-tools</a>. Install with <code>lms clone webbrain/web-tools</code> and toggle it on in any LM Studio chat β€” any tool-capable model can then call those two tools without you installing the browser extension. Pure Node, no headless browser. Source: <a href=\"https://github.com/webbrain-one/webbrain/tree/main/lmstudio-plugin\" target=\"_blank\">lmstudio-plugin/</a>.",
  "faq.scroll_during_run.q": "Can I scroll up or down while WebBrain is working on my tab?",
  "faq.scroll_during_run.a_html": "<p>Most of the time, yes β€” especially in <strong>Ask mode</strong> for read-only requests such as \"Summarize this page.\" You can continue scrolling while WebBrain works.</p><p><strong>When should I avoid it?</strong> Auto-screenshots reflect the tab's current viewport. While WebBrain is actively inspecting or interacting with visible content in <strong>Act or Dev mode</strong>, scrolling can move a target or change what the model sees. Let that step finish before scrolling; outside those moments, you can continue using the page.</p>",
  "faq.tab_switch_during_run.q": "Can I switch tabs while WebBrain is working?",
  "faq.tab_switch_during_run.a_html": "<p>Yes. As of <strong>27.0.0</strong>, WebBrain avoids stealing focus in both Chromium-based browsers and Firefox unless foreground access is strictly necessary. The run stays attached to the tab where it started, so you can safely work in another tab.</p><p>If a rare step needs the original tab in front, WebBrain may briefly focus it. Avoid clicking or typing in the same tab while WebBrain is actively acting on it.</p>",
  "faq.profile.q": "How does Profile auto-fill work, and is it safe?",
  "faq.profile.a_html": "<p>Profile auto-fill is an optional feature in <strong>Settings β†’ Memory</strong>. You enter a short bio β€” name, work email, company, and a <em>throwaway</em> password for low-stakes signups β€” and toggle it on. When enabled, WebBrain appends that text to the agent's system prompt so it can fill signup forms without asking every time.</p><p>The text is stored <strong>in plaintext</strong> in your browser's local storage. It is <strong>not</strong> transmitted to the WebBrain project, but it <strong>is</strong> sent to whichever LLM provider you have configured on every turn, as part of the system prompt. Off by default.</p><p><strong>Do not put passwords for important accounts</strong> (Google, Apple, iCloud, banking, work SSO, primary email) here. Those accounts should use 2FA and shouldn't be handed to an agent anyway. A throwaway password you reuse for newsletter signups and free trials is the intended use case.</p>",
  "faq.screenshot_redaction.q": "Does the Screenshot redaction setting guarantee my private data never reaches a cloud vision model?",
  "faq.screenshot_redaction.a_html": "<p>No β€” it reduces exposure, it is not a guarantee. When enabled (<strong>Settings β†’ Advanced β†’ Screenshot redaction</strong>, off by default), WebBrain pixelates form fields and text that looks like an email address or phone number on every screenshot <em>before</em> it is sent to a vision model. Detection runs entirely on your device using DOM heuristics; nothing extra is transmitted.</p><p><strong>It is best-effort and fail-open.</strong> If the detector cannot run on a page β€” for example immediately after a navigation, on PDF viewers, or on restricted browser pages β€” the screenshot is still sent <em>unredacted</em> rather than aborting your task. Content the DOM heuristics cannot see (text drawn on a canvas, PII inside images, unusual widgets) may also slip through. And the setting only covers screenshots: page <em>text</em> sent to the model is not redacted by it.</p><p><strong>For full vision privacy, use a local model.</strong> With an offline provider such as llama.cpp or Ollama, screenshots and page text never leave your machine, so there is nothing to redact in the first place β€” that is the only configuration where privacy is guaranteed. See \"Can I use WebBrain completely offline?\" above.</p><p><strong>Don't have a machine that can run a vision-capable local model?</strong> A good middle ground is a small local model for planning and tool calls (fast and light enough for a modest CPU/GPU) paired with a cloud provider for the vision sub-calls, with Screenshot redaction turned on. You still keep your conversation and page text on-device, and every screenshot that does leave for cloud vision has form fields and detected emails/phones blurred first β€” reduced cloud exposure without needing hardware for a full local vision model. Configure the vision split under <strong>Providers β†’ Vision</strong> (see \"How does WebBrain keep cloud LLM bills under control?\" above).</p>",
  "faq.cookies_paywalls.q": "What does WebBrain do with cookie banners and paywalls?",
  "faq.cookies_paywalls.a_html": "<p><strong>Cookie banners:</strong> WebBrain recognizes consent banners from common frameworks (OneTrust, Cookiebot, Didomi, Quantcast, Google Funding Choices, TrustArc) and dismisses them before reasoning about the page. Priority is \"Reject all\" / \"Reject non-essential\" / \"Only necessary\" when clearly visible; it falls back to \"Accept all\" rather than disappearing into the \"Manage preferences\" maze.</p><p><strong>Paywalls:</strong> WebBrain reports the paywall honestly and tells you what it could actually see (headline, dek, first paragraphs). It does <strong>not</strong> attempt to bypass paywalls β€” no archive.today, 12ft.io, cookie clearing, JS disabling, or reader-mode tricks. If you want the full article, log in with a subscription or ask WebBrain to look for free coverage of the same story.</p>",
  "faq.multilingual.q": "Does WebBrain support a dry-run mode?",
  "faq.multilingual.a_html": "<p>As of <strong>7.0.0</strong>, not yet. Dry-run mode is planned and already on the roadmap.</p>",
  "faq.page_context.q": "Does WebBrain send the whole DOM to the AI model?",
  "faq.page_context.a_html": "<p>No. WebBrain does not send raw HTML or a raw DOM dump by default. It starts with the page URL and title, matching site guidance, and an optional viewport screenshot when vision is enabled.</p><p>When a task needs page content, WebBrain reads it on demand as a reduced semantic accessibility tree or extracted text. Reads are visibility-filtered where appropriate, character-capped, and paginated. Raw page-source access is available only in Dev mode.</p>",
  "faq.offline_licensing.q": "Why is WebBrain 33.0.0 and later GPL-licensed?",
  "faq.offline_licensing.a_html": "<p>WebBrain 33.0.0 and later is distributed under GPL-3.0-or-later because the browser extension bundles and integrates the GPL-licensed Xapian/libzim WebAssembly runtime for full-text Wikipedia ZIM search.</p><p>The extension packages include the applicable GPL notices, and complete corresponding source for the bundled runtime is published alongside each release. Releases before 33.0.0 remain MIT-licensed, and independently published subprojects and third-party components retain their stated licenses. See the <a href=\"https://github.com/webbrain-one/webbrain/blob/main/docs/offline-rag-licensing.md\" target=\"_blank\">licensing decision record</a> for details.</p>",
  "faq.token_conscious.q": "How does WebBrain keep cloud LLM bills under control?",
  "faq.token_conscious.a_html": "<p>Three independent layers:</p><p><strong>Token-conscious screenshots.</strong> Before any image leaves your machine, WebBrain resizes it (shorter side capped, preserving aspect ratio) and iteratively JPEG-compresses it until it fits a per-turn image-token budget. A 2000Γ—1200 screenshot that would cost you ~1,500 input tokens on GPT-4o gets compressed down to ~300–500 tokens with no practical loss for page-reading tasks. Implemented in <code>_fitImageDimensions</code> with unit tests for the budget math.</p><p><strong>Smart context trimming.</strong> Conversation history, tool outputs, and inline DOM dumps are bounded per turn and trimmed oldest-first when the active model's context window is approaching full. You won't see a run silently balloon from 10k tokens to 100k because a <code>read_page</code> returned a novel-length article.</p><p><strong>Dedicated vision model.</strong> Pair a cheap text model (e.g. GPT-4o-mini) for planning and tool calls with a separate vision-capable model (e.g. GPT-4o) only for screenshots, so you don't pay multimodal-model prices on every single turn. Configure under <strong>Providers β†’ Vision</strong>.</p><p>Net result: long sessions with cloud providers stay predictable. For full control, use llama.cpp locally β€” zero per-token cost.</p>",
  "faq.contribute.q": "Can I contribute to WebBrain?",
  "faq.contribute.a_html": "Absolutely! WebBrain is GPL-3.0-or-later-licensed and welcomes contributions. Check out the <a href=\"https://github.com/webbrain-one/webbrain\" target=\"_blank\">GitHub repository</a> for issues, feature requests, and contribution guidelines.",
  "oss.title": "100% Open Source",
  "oss.desc": "WebBrain is GPL-3.0-or-later-licensed. Inspect the code, contribute features, or fork it and make it your own.",
  "oss.star_btn": "Star on GitHub",
  "share.title": "Spread the word, share the love",
  "share.desc": "WebBrain is GPL-3.0-or-later-licensed and runs entirely in your browser. If it's useful, drop us a star or share it β€” that's how independent open-source projects get found.",
  "share.star_btn": "⭐ Star on GitHub",
  "share.x_btn": "Share on X",
  "share.linkedin_btn": "Share on LinkedIn",
  "share.mastodon_btn": "Share on Mastodon",
  "share.bluesky_btn": "Share on Bluesky",
  "share.text": "WebBrain β€” free, open-source AI browser agent for Chrome & Firefox. Multilingual UI, token-conscious by default, GPL-3.0-or-later-licensed.",
  "social.mastodon_label": "WebBrain on Mastodon",
  "social.bluesky_label": "WebBrain on Bluesky",
  "social.discord_label": "WebBrain on Discord",
  "social.huggingface_label": "WebBrain on Hugging Face",
  "footer.tagline": "On a mission to make the web smarter.",
  "footer.copyright_html": "Β© 2026 WebBrain. Built by <a href=\"https://emresokullu.com\" target=\"_blank\">Emre Sokullu</a> &amp; <a href=\"https://github.com/webbrain-one/webbrain/graphs/contributors\" target=\"_blank\">contributors</a>.",
  "footer.link_github": "GitHub",
  "footer.link_companion": "Desktop Companion",
  "footer.link_privacy": "Privacy Policy",
  "footer.link_report_bug": "Report Bug",
  "footer.link_license": "GPL-3.0 License",
  "footer.link_press_releases": "Press",
  "footer.link_compass_subscription": "Compass Subscription",
  "features.captcha.title": "Optional CAPTCHA Solver",
  "features.captcha.desc": "Enable optional CAPTCHA providers with your own API keys and weighted fallback. hCaptcha uses NopeCHA or NoneCap; other coverage varies by provider. <a href=\"/docs/captcha/\" style=\"color:var(--accent);text-decoration:none;\">See provider coverage and setup</a>.",
  "features.captcha.note": "Off by default. BYO key β€” no CAPTCHA service is contacted unless you enable it.",
  "faq.cdp.q": "What is the \"WebBrain started debugging this browser\" banner? Why does WebBrain use CDP?",
  "faq.cdp.a_html": "To act on a page reliably, WebBrain uses the Chrome DevTools Protocol (CDP) through the standard <code>chrome.debugger</code> extension API β€” that's what triggers Chrome's \"WebBrain started debugging this browser\" banner. CDP is what lets the agent click and type with <strong>trusted input events</strong> that modern sites actually honor; synthetic events fired from a content script get rejected by many sites, Web Components, and framework-controlled inputs. It's also how WebBrain captures pixel-accurate screenshots for the vision fallback and reaches into <strong>cross-origin iframes and shadow DOM</strong> that content scripts can't see.</p><p>Read-only <strong>Ask mode does not need CDP</strong> β€” page and accessibility-tree reads go through ordinary content scripts. It's <strong>Act mode</strong> reliability and cross-origin work that require it, and in a Manifest V3 extension there is no non-debugger API for trusted input. WebBrain attaches the debugger <strong>only when an action needs it</strong>, per tab, and the whole extension is open-source so you can audit exactly what the session does. CDP is powerful, so we treat its footprint as something to contain β€” attach on demand, keep reads off it β€” and tightening that further is on our security roadmap."
}