The word "encrypted" appears on nearly every product page, and it describes at least three different arrangements that offer very different protection.
Three arrangements sharing one word
Encryption in transit protects the connection between your device and the server, so somebody on the same network cannot read what passes between them. Encryption at rest means the data is encrypted on the disk where it is stored, so a stolen drive is not a readable drive. Neither protects your data from the operator of the service, because the operator has to be able to read it to do their job. In both cases the operator holds the keys.
End-to-end encryption means something else. Data is encrypted before it leaves your device and decrypted only on the devices of the people you shared it with. The server stores ciphertext it cannot read. When a page says "encrypted" and stops there, it almost always means the first two.
Who holds the keys
The question that separates a claim from a fact is who can decrypt.
A useful test involves your password. If a service can reset your password and restore your access, then a path to your data exists that does not run through your passphrase. Whoever operates that path can read your data, whether or not they do today. For many products that is a reasonable arrangement. It is not end-to-end encryption in any meaningful sense.
A second test involves where the key is made. If your passphrase is sent to a server at sign-in, or the key is derived there, the server has what it needs. If the key is derived on your device and only a value that cannot be reversed is transmitted, it does not. CamyCal derives keys from the passphrase on the user's own device using Argon2id, never sends the passphrase, and stores a hash that can check a sign-in but cannot be turned back into the key.
The only question that decides anything
Before comparing tools, write one sentence: who am I protecting this from.
Protecting files from a stolen laptop is a different problem from protecting them from the company storing them. Protecting a calendar from a stranger is different again from protecting it from a family member who shares the account, or from a demand served on the operator. Encryption at rest is a complete answer to the stolen drive and no answer at all to the operator. End-to-end encryption is a real answer to the operator and no answer to somebody holding your unlocked phone.
A threat model is not a document a security team writes. It is the sentence that names your adversary, and it is the only thing that makes a comparison mean anything.
Why every web app asks for trust
One limit applies to every browser-based tool. A web app is code the operator sends you. Each time you load the page, you run the software they chose to serve that day.
End-to-end encryption limits what the operator can extract from their database. It does not limit what the operator can do with the code they hand your browser. A modified version could send your passphrase somewhere else, and nothing in the design would stop it. That is a property of web-delivered software in general, not of any one product.
The only complete answer is a client you can inspect and build yourself, or a native application compiled from source you have read. CamyCal is a web app, so using it means trusting the operator not to serve modified code. That trust can be reduced, by keeping the codebase small and loading nothing from third parties, but it cannot be eliminated.
What evidence is worth looking for
Claims are cheap, so look for the things that would be inconvenient to fake.
- A published description of what the operator can and cannot see. Vague answers are informative.
- Whether the app loads third-party scripts. Encryption code sitting next to an analytics tag is only as private as the tag.
- What happens when a password is lost. If losing it does not lose the data, the operator has a way in.
- Whether encryption happens on the device or on the server.
- Whether the claim is stated in a form you could check.
CamyCal loads no third-party scripts, no analytics and no web fonts, and sets one strictly necessary session cookie. It is free during early access, with a planned price of US$2.95 a month or US$0.99 a month billed yearly.
Pick the threat you actually care about, then pick the tool whose design makes that threat impossible rather than merely discouraged. Ask what the operator can see, and how you would know. If the answer is vague, that is the answer.