Clear answers before you trust us.
The practical questions: what enters, what leaves, how the beta works, and how your data stays under your control. If we didn’t cover it, ask; the honest answer is more useful than a vague one.
What is The Wheel?
The Wheel is where a company builds a record it owns: notes, documents, conversations, decisions, and sources. It is the private layer between that record and the AI you choose to use. It is built around one rule: your notes and uploads are encrypted in your browser before they reach our servers, and AI gets access only with your permission, which you can take back.
How is this different from other AI tools?
Most AI tools tie your context to the tool itself. Some keep it in the model’s memory, some in their workspace, and either way what you’ve accumulated becomes the reason you can’t leave. The Wheel is the opposite bet: a knowledge layer that belongs to you, usable with whatever tool you work in. Your record compounds in one private place instead of being re-explained every session, and when an external model helps, it works only with what you allow, only until you take that permission back.
Can I use other AI tools with The Wheel?
Yes. The Wheel ships a published MCP server (@jointhewheel/mcp-server) that works with Claude, Cursor, and other MCP-enabled tools. We name the ones we have verified. Those tools can search the parts of your knowledge you allow, list content, look up entity info, and save their summaries back to your record. The same permissions and receipts apply as everywhere else. You can also export your data to take it elsewhere.
Is it as capable as ChatGPT or Claude?
The Wheel is not trying to be a general chatbot. It uses a mix of frontier and strong open-source models, tuned for knowledge work against your own corpus: recall, synthesis, and source-grounded answers. For coding or design execution, a specialized tool is usually better, so The Wheel brings your context to those tools through its MCP server instead of becoming every tool itself.
Is there a team version?
Yes. Each person has a private space only they can see, and the team has a shared space. The boundary between them is a promote step you control: private notes stay private until you deliberately share them.
Does my data ever leave The Wheel?
Only when you ask AI to work with your content, and only with your permission. Each permission is limited in time, covers only what you allow, and can be taken back. Your receipts page shows every permission and what happened: what entered, what stayed private, what left, and what came back.
How is my data protected?
Your notes and uploads are encrypted in your browser before they reach our servers. No raw key is stored anywhere, on any side; we store ciphertext. When you ask AI to work with your record, it gets access only with your permission, only to what you allow, and only until that permission expires or you take it back. When you search, your query is processed on our servers within a bounded window and is not retained after results are returned. That window is physical, not a policy: the key lives only in server memory while you work, it is wiped when the session closes, and if our servers restarted mid-session the key would simply be gone until your browser provided it again. Sign-in is passwordless: you use a passkey.
Where is my data stored?
Content objects are primarily stored in Google Cloud Storage, with metadata and vector-search data in PostgreSQL on Google Cloud Platform. Both are US-hosted. A small amount of server-side data is stored unencrypted so the system can function: your email address, account preferences, subscription status, and basic usage metadata. None of your actual content is included. If data residency is a requirement for your organization, reach out.
What logs do you keep?
We avoid storing raw content in operational logs: notes, documents, and query text are not intended to be written into standard log streams. We do log metadata: timestamps, request identifiers, performance metrics, and error diagnostics. Some features also store interaction data in product databases where needed to operate them. Security records, like the device and IP address on your sign-ins, are tied to your account on purpose: their whole job is to help spot a session that isn’t you.
For external AI queries, what logs the provider keeps is governed by their policies. That is why sending anything outside is explicit, happens only with your permission, and shows up on your receipts page.
Do you sell my data, run ads, or train on my content?
No, no, and no. There is no advertising anywhere in the product or on this site: no ad pixels, no tracking SDKs, no behavioral analytics. Site analytics is cookieless and aggregate. We do not sell your data, we do not share it for advertising, and when an external model is used with your permission, we use API tiers where your content is not used for training.
Here is what we do watch. Anonymous, aggregate analytics: page views and a small set of named events, counted by a cookieless tool that never receives your email, your content, or any identifier pointing back to you. And when something breaks, an error report, scrubbed before it leaves the app: no message bodies, no cookies, no query text.
If that ever changes, it will happen one way: we ask you first, plainly, and you say yes or no. Never a quiet terms update.
We also think the training value of your data belongs to you. You can export your record at any time, and it comes out in structured, machine-readable formats, so if anyone gets to train a model on your record, it’s you.
Can I export or delete my data? What if The Wheel is acquired?
Your data belongs to you. The Wheel supports export and deletion controls for key data types today, including account-level privacy export and deletion flows, and you can delete content on demand, including before any hypothetical change of control or terms change. We document portability and deletion plainly, and update those documents transparently if anything changes.
Can I use this for healthcare data?
For personal use, yes: you can upload your own health information, and it follows the same encrypted architecture as other personal content. Note that your own health records are not automatically governed by HIPAA, which applies to covered entities and business associates, not individuals managing their own records.
For healthcare providers: we do not currently recommend providers use The Wheel for patient data. We have not completed a formal HIPAA compliance audit and do not yet offer Business Associate Agreements. If you’re a healthcare organization interested in The Wheel, contact us.
Which connectors work today?
Google Drive is live in the beta; you pick the folders it can see. Gmail, Google Calendar, Slack, Notion, and GitHub are built and in final testing, and they open as testing completes. Linear and Jira are coming soon. Direct upload works for anything that isn’t connected yet.
How does The Wheel’s own chat work?
You ask in plain language. Search finds the relevant pieces of your record, and only those pieces are opened, inside the bounded processing window, to build the answer. The answer comes back with its sources shown. Underneath, a knowledge graph links your notes, documents, and conversations, so search can follow connections instead of just matching words: ask about a decision and the meeting note that led to it can surface too.
What AI models does The Wheel use?
A mix of frontier models and strong open-source models, chosen for the task. The default path is self-hosted: everyday chat, search, and the systems that screen content for personal details all run on models we operate inside The Wheel’s own cloud, and an outside provider is used only when you allow it. The initial beta runs a fixed, tested set, and your receipt names the provider each request went to, when, and under which permission; over time, routing expands so every question is matched to the model that fits it best. Models are swappable: the knowledge layer stays with you.
What is the receipts page?
Your receipts page shows every permission you’ve given: what entered, what stayed private, what left, and what came back. You can inspect each one and take it back, and every run carries a full four-layer trace: result, trace, custody, what’s next. An empty record is never presented as proof that nothing left.
How does the MCP server work?
Install @jointhewheel/mcp-server in Claude, Cursor, or another supported MCP client, and that tool can search the parts of your knowledge you allow, list content, look up entity info, and save its summaries back to your record. It sees only what you let it see, and you can cut off its access the moment you change your mind. We don’t promise every MCP host will work; clients vary, and we name the ones we’ve verified.
How do I get access?
The beta is invite-only and free. Request access below and we’ll be in touch as spots open. If your invitation isn’t working, email info@jointhewheel.com.
How long does it take to set up?
The goal is that you can upload a first document and start asking useful questions in one sitting: no DevOps, no hardware, no infrastructure to manage. A larger team archive takes longer, mostly because deciding spaces, sources, and permissions deserves real thought.
What should I expect from the beta?
A product that is still being shaped, described honestly. Core knowledge and privacy work matters more to us than pretending every edge is finished, and your feedback shapes the roadmap. People who request beta access go first as new capabilities finish testing.
Still have a question?
Email us and a person answers. The beta is invite-only and free.