Tinfoil
Tinfoil Inc. · Chatbot
Tinfoil is a private AI chat app that keeps your conversations confidential by running every prompt inside secure hardware enclaves. It's also an encrypted AI chat tool that works in the browser and on iOS, costs $20/month, and lets you verify the privacy claims yourself instead of taking them on faith. You chat with powerful open-source and multi-modal models, upload documents, search the web, and sync encrypted backups across devices. No cloud provider gets a readable copy of your data.
About Tinfoil
What Is Tinfoil
Tinfoil is a private AI platform from Tinfoil Inc., a San Francisco company founded by researchers with backgrounds at MIT, Cloudflare, and NVIDIA. Its flagship product, Private Chat, is a private AI assistant that runs your prompts inside secure hardware enclaves, a setup also called a trusted execution environment (TEE). The enclave is a sealed part of the chip that processes your request. Not even Tinfoil's own engineers can look inside it, and the platform publishes the code running there so you can check it.
The problem it solves is simple. Most AI chatbots read your messages to train models or improve service. You have no way to prove otherwise. Tinfoil flips that. Your prompts stay encrypted, backups are stored as ciphertext, and the app verifies the enclave's integrity right in your browser before you type anything. That's secure enclave AI you can actually confirm, not just read about.
The tradeoff is cost and scope. There's no free tier, so you pay $20/month from day one, and the model lineup is open-source rather than the newest closed models from OpenAI or Google. Want the cutting edge of raw capability? That's not this. Care more about keeping sensitive questions out of anyone else's hands? Then it's exactly this.
Getting Started
- Open chat.tinfoil.sh in your browser or install the iOS app. No credit card is needed to look around first.
- Sign up and subscribe to the Private Chat plan at $20/month (billing is monthly, cancel any time).
- When the verification panel loads, confirm the enclave attestation passed. This is the step that proves your session is actually private.
- Start a chat, pick a model, and send a prompt. Set up a passkey if you want your encrypted history to sync to another device.
- Upload a document or turn on web search when you need outside context inside the same private workspace.
Product Information
A quick look at Tinfoil's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Privacy-conscious professionals
- Journalists and researchers
- Developers exploring private inference
Tasks
- Confidential drafting
- Document analysis
- Sensitive brainstorming
Scenarios
- Checking a worrying symptom or medication question before a doctor visit, without that query becoming ad data.
- Reviewing an employment contract or lease and wanting a plain-language summary without uploading it to a service that stores it.
- Building a small internal tool that needs private AI inference and an OpenAI-compatible endpoint.
Key features
Secure Enclave Processing
Every conversation runs inside a hardware enclave, a locked region of the processor that isolates your request from the rest of the machine. Tinfoil's engineers get the same restricted view you do. That's the whole point. The company calls this provable zero data retention, and it covers the inference itself, not just the storage layer.
In-Browser Attestation
Before you send your first message, the app checks that the enclave is running the exact open-source code Tinfoil published. This verification happens in your browser. So you're confirming the claim on your own machine rather than trusting a marketing page. Because the request never leaves that sealed environment in a readable form, there's no log a subpoena could pull and no training pipeline that could quietly absorb what you typed.
Projects
Projects group your chats, documents, and context into separate workspaces with isolated memory. Ask about a work contract in one project and a personal topic in another. Neither bleeds into the other's context. It keeps long-running threads from turning into one giant pile of unrelated history.
Multi-Modal Document Processing
You can upload documents and have the models extract, summarize, and reason across them. Because this happens inside the enclave, the content never gets handed to a separate parsing service that could retain a copy on its own servers. For anyone working with sensitive source material, that's the difference between usable and off-limits.
Private Web Search
Web search pulls fresh information into your workspace without tying the query to your identity. You get current answers when the model's training data isn't enough. And you don't leak your reading habits in the process. It's a small thing that matters a lot for research on sensitive topics.
Encrypted Sync With Passkeys
Tinfoil stores encrypted backups so your history moves between your phone and browser. Passkeys handle moving your decryption keys between devices, which means Tinfoil can sync the ciphertext across your phone and browser but still can't read the messages inside it. Setup is a few taps, not a key-management project.
OpenAI-Compatible API
The inference API follows OpenAI's format, so existing code needs only minimal changes to switch over. It's open source and ships with SDKs that verify attestation automatically on each connection. Teams that want private inference without rewriting their stack get an easy path.
Pros and cons
Pros
- Verifiable privacy: the app checks enclave integrity in your browser, so the claim is testable, not just stated.
- Broad capabilities on the paid plan, including document upload, web search, speech to text, and multi-modal models.
- Cross-device sync stores only ciphertext, so backups don't create a readable copy of your history.
- OpenAI-compatible API makes it easy to route existing code to private inference.
- Open-source security-critical code lets outside auditors check how it works.
Cons
- No free tier. You pay $20/month before you can use chat at all, which is steep next to free competitors.
- The model lineup is open-source. If you need the newest flagship closed models, they aren't here.
- Usage is capped at roughly 20M input and 1M output tokens per hour under a fair-use policy, and bursts can get throttled.
- Desktop coverage is browser-only for now, with iOS as the only native app.
Frequently asked questions
It's built to be checkable, which is unusual. Your prompts run inside a hardware enclave, and the app verifies that enclave's integrity in your browser before you chat. Tinfoil publishes its security-critical code on GitHub, so the claim rests on code you or an auditor can inspect rather than a promise in a policy page.
Related content
Explore related tools, skills, and articles for Tinfoil.
Tinfoil Alternatives
Knocket
Knocket (Tencent RTC team) · Productivity · Business · ChatbotKnocket is a free live chat and contact page tool that lets visitors message you from your website, app, or a shareable link, with every reply handled from Telegram or one inbox.
Agibot
Agibot · Productivity · AI Assistant · ChatbotAgibot is a personal AI agent for Mac that keeps running after you close the window, runs routines on a schedule, and works from Telegram or Slack with your own API keys.
Melete
Melete · Productivity · AI Assistant · ChatbotMelete is an AI workspace for teams where agents research, draft, and act in Slack, Notion, Linear, and GitHub. Shared memory and approval gates keep people in control.