Secret Chat AI — Private AI Chat Sidebar 作者： Opossum DOO

Ask ChatGPT, Claude, Gemini, Grok, DeepSeek and Perplexity from one sidebar, without signing your name to a single question. No profile, no server-side chat archive — history stays in your browser. It cannot read your tabs. Free daily, no account.

<b>THE BIG AI APPS KNOW WHO YOU ARE</b>

Every time you open ChatGPT, Claude or Gemini, you sign in first — and that is the whole design. The conversation is stored, tied to your identity, and folded into a profile that keeps growing with every question you would not ask out loud: a symptom, a salary, a resignation letter, a legal worry, a business idea you are not ready to defend. The better these models get, the more you tell them — and the more there is on file about you.

Secret Chat AI is built to break that trade. You get both halves of what you actually want: the strongest AI models in the world, asked anonymously, in a sidebar that sits beside whatever you are already doing. No profile. No server-side chat archive. No identity attached to a single prompt.

<b>THE EXTENSION IS FREE, AND NOTHING IS SOLD HERE</b>

This extension is a free client for Secret Chat AI, a web service that runs at secret-chat.ai. Installing it costs nothing, and there are no purchases, subscriptions, payments, upgrades or unlocks inside the extension or through Firefox Add-ons. It opens the web app in the browser side panel — that is the whole of what it does.

The service itself has a free daily quota that needs no account and no payment method, so you can install this and start asking questions today without creating either. If you later want more than the daily quota, credits are bought on our website — not in this extension, and not through this store.

<b>ONE SIDEBAR INSTEAD OF SIX TABS</b>

An AI assistant is only useful at the moment a question appears — while you are reading, writing, comparing, coding, studying, deciding. Secret Chat AI lives in the browser side panel, so the chat is one click away and stays open while you move between tabs: ask, read the answer, keep working. No switching to another app, no copying your question into a separate window, no losing your place.

And because it is an aggregator, this one panel replaces not one AI tab but all of them. ChatGPT, Claude, Gemini, Grok, DeepSeek and Perplexity live in a single interface with a single balance — instead of six logins, six subscriptions and six browser tabs. Ask on the free model first; branch the same conversation into a stronger one when the question turns out to deserve it.

<b>EVERY FRONTIER MODEL, ONE PANEL, ONE BALANCE</b>

<b>CHATGPT</b> — the all-rounder. Three tiers, from instant answers to a flagship that handles genuinely hard reasoning, plus web search and image generation. If you do not know which model suits a question, this is the one that will not disappoint you.

<b>CLAUDE</b> — the writer and the thinker. The strongest of the lineup for long-form text, careful step-by-step reasoning, code and large documents, and it reads big PDFs without losing the thread. It also carries a second flagship tuned specifically for creative and narrative writing, so you can choose between rigour and voice.

<b>GEMINI</b> — fast, and grounded in Google's index. The quickest of the smart families for everyday questions, with web search when a question needs sources, a lightweight tier for instant replies, and the ability to draw and edit images right in the conversation.

<b>GROK</b> — the one that knows what happened this morning. It searches the live web and X, which no other model here can read, so it is the one to ask about breaking news, a running public argument, or anything where "current" matters more than depth. It draws images too.

<b>DEEPSEEK USA</b> — free, every day, every tier. An open-weights model served exclusively through an allowlist of US-headquartered providers that declare they do not collect user data — never through DeepSeek's own Chinese API. It is text-only and does not search the web, and it is the app's default because for most questions it costs you nothing.

<b>PERPLEXITY</b> — search first, answer second. Every reply is built from live web sources and cites them, which makes it the one for research, fact-checking and "is this actually true", rather than for conversation. Also free every day.

<b>IMAGE GENERATOR</b> — six image models in one tool. Aspect ratio, output resolution, number of variants, style presets and negative prompts, each exposed where the model actually supports it. Follow-ups understand context: generate a poster, then say "make it blue" or "same scene at night", and the edit lands on the picture you meant. Attach your own photo and it becomes one-click image-to-image editing.

Always the current releases. Every model here is whatever its provider is shipping right now — when a new one lands it appears in the picker on its own, and you do not pay for six subscriptions to keep up.

<b>SMART AGENTIC MODE — CHOOSE A MODEL, NOT A SETTING</b>

Every family above also has an agentic mode, and it is the default. You pick ChatGPT, or Claude, or Grok — and its router reads your actual question and decides the rest: how much thinking it deserves, whether to go and search the web, whether to draw an image, and which tier of that family should answer. A quick question gets a quick, cheap answer; a hard one gets the flagship. You watch what it decided in a live status line while it works, written in your own language.

The point is that you stop paying flagship prices for "what time is it in Tokyo", and stop getting a rushed answer to something that deserved research — without having to learn which tier is which. If you would rather decide yourself, every family also exposes its tools directly: Fast Chat, Web Search, Deep Research and Create Image.

<b>WHAT "PRIVATE" MEANS HERE, EXACTLY</b>

Secret Chat AI is an anonymizer. Precisely:

<ul>
<li>No profile is built about you. Sign-up, if you ever want it, is email-only — no name, no phone number, no behavioural tracking, no ad pixels.</li>
<li>No conversation is ever associated with you.</li>
<li>Your queries reach the AI providers anonymously — no identity, no account and no IP of yours attached.</li>
<li>Your queries are never used to train AI models. Every provider is accessed under business API terms with training opted out.</li>
<li>Chat history lives in your own browser (IndexedDB and OPFS), not on our servers. There is no server-side archive to leak, to sell, or to hand over — a request for your chat history cannot be satisfied, because the history does not exist on our side.</li>
<li>Every answer generates a Session Privacy Report, a PDF naming the model and the provider and recording whether a deletion request was sent, was unnecessary because nothing was stored, or is not offered by that provider. You do not have to trust a settings toggle; you get a receipt.</li>
<li>DeepSeek runs only through US-headquartered providers that declare they do not collect user data — never through DeepSeek's own Chinese API.</li>
</ul>

<b>AND WHAT IT IS NOT, SAID PLAINLY</b>

A model has to read your message in order to answer it. Your text is not encrypted or redacted on the way to the provider, and no privacy tool can change that without breaking the answer. Removing names, numbers and identifiers before you send is your own responsibility — the app can warn you (see below), but the judgement is yours. Secret Chat AI removes YOU from your queries; it does not remove the data from your messages.

<b>IT CANNOT READ THE PAGE YOU ARE ON</b>

Most AI sidebars ask for permission to read and modify every website you visit — that is what "summarize this page" costs. This extension asks for two domains: its own app and its own API. There is no content script, nothing is injected into any page, and nothing from the sites you browse passes through this extension. Check the permissions prompt when you install — that short list is the entire ask.

The honest consequence: it will not summarize your current tab automatically, because it cannot see your current tab. Paste the text or attach the PDF instead — the models read both. That is the deliberate trade: a sidebar that cannot watch your browsing is a sidebar you do not have to trust about your browsing.

<b>WHAT PEOPLE ACTUALLY USE IT FOR</b>

<ul>
<li>Writing they would rather keep out of an account history: a resignation letter, a complaint, a salary negotiation, a difficult reply that took four drafts.</li>
<li>Health questions, asked in plain words, without the question becoming part of a permanent profile.</li>
<li>Everyday work: summarize a long PDF, rewrite a paragraph until it sounds human, translate a document, draft the email you have been postponing, untangle a spreadsheet formula.</li>
<li>Code: explain an error message, review a function, write a script, compare two approaches — with a model strong enough to argue back.</li>
<li>Study and research: cited answers from live sources, deep research reports on the flagship tiers, follow-up questions until it actually makes sense.</li>
<li>Second opinions: run the same question through two different models in branched threads and compare what comes back.</li>
<li>Images: posters, product shots, illustrations, avatars — generated, then refined by conversation instead of re-prompting from scratch.</li>
</ul>

<b>YOUR HISTORY LIVES WITH YOU</b>

Conversations are stored in your own browser, and export is built in: one chat, one project, or everything, as a ZIP you can import into any other browser — or keep as a backup. Nothing syncs through a cloud, because there is no cloud copy to sync. That is a real trade (moving devices is a manual export and import), and we chose it on purpose: the only place your history exists is the place you control.

<b>EVERYTHING ELSE THE APP DOES</b>

<ul>
<li>Projects — group related chats, give each project its own standing instructions.</li>
<li>Threads — branch any message into a nested conversation that inherits the context up to that point; branch the branch if the tangent grows its own tangent.</li>
<li>Message-level control — delete a single message, or a single image, from a chat. The big AI apps only let you delete a whole conversation.</li>
<li>Message actions — copy, retry, or edit your message and get the reworded version as a new message, with the original kept.</li>
<li>Custom instructions — global rules applied to every model, plus per-project ones.</li>
<li>Attachments — images and PDFs, with the attach button offering only what the selected model can actually read.</li>
<li>Voice input — dictate, get a transcript in the message box, edit it before sending.</li>
<li>A system notification when a slow answer is ready — it fires only while the app is in the background and never contains chat content.</li>
<li>An optional local check that warns before you send something that looks like personal data. A heuristic, not a guarantee — a seatbelt, not an autopilot.</li>
<li>Dark mode and eighteen accent colours, stored locally, never attached to an account.</li>
</ul>

<b>FREE TO START, FAIR AFTER THAT</b>

The free daily quota renews every day and requires nothing from you — no account, no card, no trial countdown. Past it, the service is usage-based: you buy credits on our website and they are spent per request, on any model, from one balance. One-time packs never expire, discounted subscriptions exist for regular use, and a reminder email goes out days before any renewal charge — no surprise billing, ever. Whatever you choose, it happens on the website: this extension stays free and sells nothing.

<b>QUESTIONS PEOPLE ASK BEFORE INSTALLING</b>

<b>Is it really free?</b> — The extension: yes, entirely. The service: a free daily quota on two models with no account needed; more than that is paid credits, bought on our website, never in here.

<b>Do I need an account?</b> — No. Try everything on the daily quota first. Registration is email-only and exists so a balance can follow you, not so we can know you.

<b>Can it summarize the page I am reading?</b> — Not automatically: it has no access to your tabs, which is the point. Paste the text or attach the file and any model will summarize it.

<b>Where are my chats stored?</b> — In your browser, on your device. Export them as a ZIP any time. We keep no copy.

<b>Which model should I start with?</b> — The default free one, in agentic mode. It answers most everyday questions well, costs nothing, and the moment you need more, the stronger families are one click away in the same picker.