The short version: compare the price per message you are allowed to send, not the headline monthly price, and read what happens when you reach the limit.
Four things, in roughly this order of size for a small service.
1. Inference. Every reply is a call to a language model, and the model is billed by the amount of text going in and coming out. Going in is not just your last message: it is the character description, whatever the service remembers about you, and the recent conversation. That is why a companion that remembers a lot costs more per reply than a one-shot question, and why some services quietly shorten what they remember when they are busy.
2. Payment processing. Card processing on a small subscription is a percentage plus a flat fee per charge. The flat part does not care that the charge is small, so on a single-digit monthly price it is a real slice, and it is charged again every month the subscription renews.
3. Tax. A seller in the EU charges VAT to consumers at the rate of the buyer's own country, so the same subscription nets a different amount depending on where you live. Check whether the price you are quoted includes VAT or adds it at checkout; both are legal and they are not the same number.
4. The heaviest user. This is the one that decides the design of the whole plan. Usage is not evenly spread: most people send a handful of messages a day and a few send hundreds. If the plan promises no limit, the price has to survive the heaviest user, or the service has to find a way to slow them down.
An unlimited plan on a metered cost has exactly four exits, and every service takes at least one:
None of those are dishonest by themselves. The problem is that they are invisible, so you cannot tell in advance which one you have bought. A stated cap is worse marketing and better information: you know what you get, and you can divide.
Take the monthly price and divide it by the number of messages the plan actually allows. That gives a cost per message you can compare across services. If a plan states no number, you cannot do the division, and that in itself is the answer to how comparable it is.
Then ask the question the number does not answer: what happens at the limit? There are only three possibilities, and they are very different for your bank account. The service charges you more per message beyond the limit; or it drops you to a cheaper model and keeps going; or it stops until the period resets. Only the third one is knowable in advance.
A free companion still pays for inference, so something else is covering it. That may be an ad-funded model, a paid tier you are being funnelled towards, or permission to train on what you type. Any of those can be a fair trade — as long as you know which one you agreed to. The place to look is the section of the privacy policy about how your content is used, not the marketing page.
Everything above applies to any service, including this one, so here are the same answers for the companion on this site, in the same terms.
What it is, and what it is not
Write to hello@toolkitlabs.org. It is a real inbox on this domain and the messages are read.