Ask us anything.
What Thaura is, how it works, and what we stand for. Straight answers, no runaround.

About Thaura
Thaura (ثورة, Arabic for "revolution") is an ethical AI companion, born from a revolution, not a boardroom. Its founders, two brothers from Syria, grew up watching the powerful decide who gets heard, who gets watched and who gets left out. When they looked for an AI that didn't work against the people using it, they found none. So they built Thaura: independent, with no ads and no investors, never trained on your conversations, and bound by a public Constitution. An AI that answers to you, not to the powerful.
It does the work, not just the talking. Every chat has its own secure computer where Thaura searches the web and reads pages, runs code, installs what it needs and makes real files: Word documents, spreadsheets, slide decks, PDFs, charts and websites you can preview beside the chat. It connects to apps like Notion, Linear and Todoist, remembers what matters across chats, keeps related chats and files together in projects, and, with the desktop app, works on your own computer.
Thaura is ethical in three key ways: • Data Privacy: We don't mine your data or spy on you • Truthful Stance: No watering down the truth. We don't dodge hard topics, and we don't pretend honest disagreements are settled either • CO₂: on average more than 200 times less CO₂ than other AI models, for the same work, reasoning included.
We build trust through complete transparency. We're not a faceless corporation, just two brothers who started this from scratch. We rejected investor funding to prevent shareholders from influencing our mission, and we're completely open about our tech stack and current limitations. Throughout this FAQ, you'll see our honest approach to where we are now and where we're headed. If you have more questions after reading through, please ask. We're committed to being as open as possible about this journey we're on together.
Technology & performance
No. Thaura is fully independent. It runs on GLM-5.3-Flash, an open-weights model published by Z.ai (Zhipu AI) under the MIT license. We run it ourselves on GPU servers we rent in Helsinki, and the Thaura Constitution guides it through its instructions. On top of that come our own safeguards: in “Ask first” mode Thaura asks before risky actions, its own computer can only reach package sources and GitHub, and it only opens links from your conversation or its search results.
Thaura runs on GLM-5.3-Flash, an open-weights mixture-of-experts model from Z.ai (Zhipu AI), released under the MIT license with native vision support. Of its 320 billion parameters, only about 18 billion are active for each token. We run it ourselves on GPU servers we rent in Helsinki, Finland, and no data goes to Z.ai. We haven't fine-tuned it: the Thaura Constitution guides it through its instructions.
Thaura handles most everyday work well: writing, research, analysis, code and files. The largest frontier models still do better on the hardest tasks. It runs on GLM-5.3-Flash, a mixture-of-experts model: only about 18 billion of its 320 billion parameters work on each token, so each answer takes a fraction of the computing power a model of its size would otherwise need. You can see where it stands in independent benchmarks (artificialanalysis.ai/leaderboards/models). Our goal isn't to achieve AGI (Artificial General Intelligence). We want to create an AI platform people can actually trust, one that puts them first.
Currently we don't have the financial resources to build our own model. When we grow and can afford it, we're definitely planning to train our own models the right way: ethically and responsibly. For now, we're focused on building something better than what's out there.
All our servers are in the EU: Hetzner in Helsinki, Finland, runs the app, each chat's computer and file storage; our encrypted backups are kept on a Hetzner server in Falkenstein, Germany; and the AI model runs on GPU servers we rent from Verda in Helsinki and operate ourselves. A few services outside those servers do specific jobs: Brave (USA) runs web searches and receives the search words, but no account details or IP address; Scaleway (Paris, France) turns voice dictation into text. Running the main model ourselves means your conversations aren't sent to another AI company to be answered.
Thaura reads your whole conversation, its files and your new message together, and that has a length limit. When a conversation grows long, Thaura summarizes its earlier part automatically so you can keep going. A single message, with the files attached to it, still has to fit on its own. If it is too long, shorten it, attach fewer or smaller files, or start a new conversation.
Privacy & security
We don't sell your data, share it with advertisers, or use it to train AI models. Your conversations, files, memories, projects and connector sign-ins are encrypted by our app (AES-256-GCM) where we store them, and the working copy of a file on a chat's own computer is protected by that server's disk encryption. The disks of our app, chat and storage servers are encrypted, the GPU servers that run the model keep no conversations on disk, data between your device and us travels over TLS 1.2 or newer, and inside our private network files move already encrypted. Our team doesn't read your chats: our software decrypts them only to answer you. The only exceptions are an answer you send us with feedback, and a page you shared that someone reports to us. There are no ads and no third-party trackers. We do count how Thaura is used in aggregate, and for 90 days we keep the times you open the app (and, on the Free plan, send messages), for the usage limit and for account emails, which you can turn off. Memory is on by default, and you can view it, clear it or turn it off at any time. A deleted chat or file disappears from your account immediately and is erased from our storage within 30 days. Our privacy policy lists everything we keep and for how long.
We're committed to providing truthful answers without watering down sensitive topics. The Thaura Constitution guides the model through its instructions, so it takes on the kind of questions other AI platforms tend to avoid, because we believe transparency is essential for building trust with users who deserve the full picture, not curated narratives.
CO₂
AI is already everywhere, and the electricity behind every reply causes CO₂. The companies behind other AI models publish no figures for it. We'd rather show you ours. Under the latest reply, Thaura shows about how much CO₂ the chat has saved so far, compared with doing the same work on other AI models. Tap it to see the CO₂ on both sides, and how we work it out.
On average, more than 200 times less CO₂ than other AI models, for the same work, reasoning included. Even a short reply comes to about 70 times less CO₂, and the more Thaura writes, the more it saves. Two things are behind it. Our model is a mixture of experts: of its 320 billion parameters, only about 18 billion work on each token, far fewer than the largest models are estimated to use. And it runs in Helsinki, on Finland's grid, which averaged about 57.5 g of CO₂ per kWh in 2025, against about 384 g on the US grid, where those models run (Ember). How we calculate this: our side is measured on our own servers in Helsinki, over real use in November 2026, with server overhead and cooling added at our provider's published factor. Their side is an independent estimate (EcoLogits) for the same amount of written output from today's flagship AI models of two leading AI companies, because those companies publish no figures of their own. Both sides are counted for reading the conversation and writing the reply. For the other models, reading is costed at the same share of writing that we measure on our own servers, and we assume they reuse as much of the conversation from their cache as we do. We measure again every month and whenever the model changes. Not included: training the models, making the hardware, our servers besides the GPUs, and your own device and network.
Here's what our providers say about their electricity: • Hetzner (Helsinki: the app, each chat's computer, file storage) states that the electricity for its Finnish site is covered by hydropower guarantees of origin (Vattenfall EPD Hydro, 2025–2026). • Verda (Helsinki: the GPU servers that run our model) states that its site runs on 50% hydro and 50% wind power. It has published no certificate for this. • Scaleway (Paris: voice dictation) states that its electricity is 100% wind and hydro, backed by guarantees of origin. Our CO₂ figures don't rely on any of this: we use Finland's average grid, a figure anyone can check.
Decentralization & open source
Decentralization means taking power away from a handful of corporations and distributing it across many users and communities. Right now, a handful of corporations control almost all AI through their centralized servers, data collection, and algorithms. This concentration of power means your digital life is controlled by companies that prioritize profit over people. We need to break that concentration of power and keep any one company from controlling everything. One of our long-term goals is to develop small, efficient models that can run locally, eliminating dependencies and giving users complete control.
We're providing this solution for regular people who don't want to host their own model, or aren't technical enough to. At the same time we're planning to open source Thaura one day, and we'd love for people to host it on their own as part of our decentralization mission. However, we're being thoughtful about timing because of serious security concerns. We've seen firsthand what happens when code is released before it's been properly hardened. On top of that, we're up against adversaries with significant security resources and skills, while we operate with far fewer. Releasing code that hasn't had time to be properly hardened could create vulnerabilities that could be exploited against our mission. We need to ensure our code is properly secured before opening it up, to prevent technical exploits and to keep our values from being co-opted.
We encourage everyone to self-host. Thaura is designed for people who aren't technical or don't know how to host their own models. Even then, GLM-5.3-Flash alone won't do everything Thaura does. Thaura adds a computer of its own for every chat, web search, transcription via Whisper (served by Scaleway), document reading, connectors, memory and projects, plus safeguards such as asking before risky actions. Additionally, running GLM-5.3-Flash at full context and speed takes datacenter-class GPUs, which is not realistic on most consumer hardware, while Thaura makes it accessible to regular people. We also give the model the Thaura Constitution in its instructions, so it's guided by our values rather than only its factory defaults.
Business & development
We're unfunded and bootstrapping this ourselves. We've deliberately declined investor opportunities because we refuse to have shareholders influence our mission. We believe ethical AI shouldn't come with corporate strings attached. We've been operational for about a year, building steadily with our own resources.
Our platform features a free version with usage limits (5 messages every 5 hours), a Pro version ($15/month) with far more room, and a Max version ($100/month) with five times Pro's room. Every plan is the same product; only the usage differs. For teams there's the Team plan, on one bill: each person gets a Pro seat at $15 a month or a Max seat at $100 a month (at least two people). Our API works with the official OpenAI SDKs: set your key and base URL and use the model name thaura. We charge $0.50 per million input tokens and $2.00 per million output tokens.
We deliberately set our pricing at $15 when competitors charge $20 to provide an ethical, affordable alternative. These rates represent the minimum we need to operate effectively while maintaining our independence.
Yes! You'll find our open roles, and how to apply, here: We'd also love for you to join our community, where you can connect with other supporters, stay updated on our progress, and get a feel for working together! Join us here:
You can cancel anytime. Open your account menu (your name, at the bottom of the sidebar) → Settings → Change plan, choose Cancel and confirm with Cancel plan. Telling us why is optional. Your subscription will be canceled and you'll retain access until your billing period ends. No hidden fees, no hassle.
No. On a team plan the admin sees your name, your email address, your role and when you joined, along with the team's invites and its bill. Never your chats, files, memory or usage, and there is no way for an admin to export them.
Each person on the team has a seat: a Pro seat at Pro's price ($15 a month, or $12 a month billed annually) or a Max seat at Max's price ($100 a month, or $80 a month billed annually), with at least two seats. Inviting someone adds a seat, charged now for the rest of the billing period. Removing someone frees their seat and lowers the bill, never below two seats. The admin can move someone between a Pro and a Max seat at any time. It's all on one bill, with invoices that show your business name and VAT number.
If you pay for Pro or Max yourself, it stops renewing when you join, and you keep it until the end of the period you've paid for. After that, the team pays for your Pro. Your chats, files and memory stay exactly as they are.
Features & availability
Every chat runs on a small virtual machine of its own on our servers in Europe, with 2 processors and 8 GB of memory, walled off from every other chat and every other person. That's where Thaura works: it runs code, installs packages from places like npm and PyPI, and builds your files. It sleeps a few minutes after the chat goes quiet and wakes when you send your next message. The files you attach and the files Thaura hands you stay with the chat. Its working space is temporary: after two days without use it is cleared, and your next message starts it fresh with those files in place.
Connectors let Thaura work in the apps you already use. Connect one with a click (some ask you to sign in): Notion, Todoist, Calendly, Cal.com, Trello, Airtable, Linear, Zapier, NocoDB, Jotform, Tally, Close, Kit, Readwise, Consensus, OpenAlex, Wikidata, DeepL, Frankfurter, CourtListener, Kiwi.com, Tarteel, QuranMajeed Prayer Times, Zabihah, Daleel, NewsCord, Muslim Business and Find A Protest. You can also add your own: any remote MCP server. Choose which connectors a chat may use from the + menu under the message box.
You decide. In “Ask first”, Thaura asks before it changes a file or runs a command on your computer, or acts through a connector, and you allow or deny each one. On its own computer in the chat, it works without asking. In Auto, it goes ahead on its own. You can switch at any time from the message box, and deleting files on your computer or saving a new skill always asks first.
Not yet. On iPhone and iPad, open in Safari.
Yes, for Windows, macOS (Apple Silicon), and Linux. It installs in a couple of clicks, and recent versions keep themselves up to date (a very old version needs one fresh download). The desktop app also lets Thaura work on your computer: it can read and edit your files anywhere your account can reach, run commands, and move files between your computer and the chat. In “Ask first” mode it asks before it changes a file or runs a command; in Auto mode it goes ahead without asking (deleting files or saving a new skill still asks first), so keep backups of anything important. Grab it at
Feedback is literally oxygen for this project! We don't read your chats to learn what works, so we rely on you to tell us. Please feel free to join our community where we discuss feedback, bugs, and feature requests.
Our API is built for both developers and businesses who want ethical AI that actually works. It works with the official OpenAI SDKs: set your key and base URL and use the model name thaura. You can find our API docs at To get started: 1. Log in to Thaura 2. Open your account menu (your name, at the bottom of the sidebar) → API 3. Create your own API key Then set your key and the base URL in the OpenAI SDK, and you're good to go!
Yes! We'd love for you to join our community where you can contribute, connect with other supporters, and stay updated on our progress! Join us here:
Ethics & values
Our partnerships serve our mission, not corporate interests. We're building partnerships within the ethical tech ecosystem we've connected with through Tech for Palestine. These are people and organizations that share our values. We don't partner with big players whose practices conflict with our values. For a few things we can't replace yet, we do rely on large companies, such as Stripe for payments and Google Play for our Android app, and we say so openly. Our partnerships must align with our core principles of ethics and human dignity.
Arabs get reduced to deserts and camels instead of APIs and keyboards. So we took the camel, the very thing used to shrink us, and made it the face of serious, independent technology. Reclaiming the symbol is the point. We're proud of who we are, and we won't adjust to fit anyone else's expectations.
"Thaura" means "revolution" in Arabic, a word we know well from Syria's revolution. Today, the tech ecosystem needs its own revolution, so we chose this name to represent our fight for ethical AI.
We believe AI should serve people, not exploit them. We're committed to building technology that aligns with our values of privacy, transparency, and truthfulness.
Getting started
Simply visit and start using our free tier. You can upgrade to Pro or Max whenever you want far more room. For developers, you can create an API key from your account menu → API.
We're still in the early stages and focusing on getting our finances on a sound footing first. We deliberately set our pricing at $15 when competitors charge $20 to provide an ethical, affordable alternative. Once we have more financial resources, we plan to introduce educational discounts and more flexible pricing options to make ethical AI accessible to everyone.
Get involved
The most important support you can give is using Thaura and sharing your experience with others. We'd love for you to join our community where you can contribute, connect with other supporters, and stay updated on our progress! Join us here:
Thaura represents a collective effort to reshape the tech industry. We're a very small team working incredibly hard, but we've already got thousands of users who've migrated over and are happy with the switch. Real change happens when people choose independent alternatives and build them together. We're stronger together!
Still have questions?
Ask away. A real person reads every message, and we'll get back to you within 24 hours.