Quick answers,
before you start.

Calypso is an AI confidante that reads between the lines of your real conversations. You can do three things with it: decode a screenshot or pasted conversation to see what the other person actually meant; talk it through with one of four AI personas when you need somewhere to put it; and switch personas depending on what you need that day. Aria is free to start, with no app to download and no appointment to book.

Yes — you can start without paying anything. The free tier gives you Aria for chat and one decode a week with a free account, which is enough to get a real feel for how it works. Pro adds Markus, Rhea and Luna, unlimited chats and unlimited decodes, and turns on the synastry (star-chart) read inside the decode report. There's no free trial trick and no credit card needed to start.

No, not to start talking. You can chat with Aria, our free persona, straight away with a limited number of free messages each day. You'll need a free account to decode a conversation, to keep your chats and come back to them, and to unlock more messages. Creating one takes an email address and a name, which doesn't have to be your real one. There's no phone number to give and no social login.

Calypso runs in your browser. There's no app to download. It works on phones, tablets, and laptops; you can pick it up on one device and continue on another with the same account.

No. Calypso is for adults aged 18 and older, everywhere. We ask you to confirm you're 18 or older when you sign up, we don't knowingly accept anyone under 18, and if we learn an account belongs to a minor we close it and delete its data. If you're under 18 and going through something hard, please talk to a trusted adult — and if you are in any kind of crisis, the resources on our Safety page, including 988 in the US and Canada, will get you to a real person who can help.

Ka-LIP-so. Same as the sea nymph in Homer's Odyssey, and the same as the music. We picked the name because Calypso, in the myth, is the listener on the shore — the one who hears the whole story when someone washes up on her island. That felt right.

Usually for a boring reason: they're busy, they saw it at a bad moment, or they're not sure what to say. One slow reply rarely means much on its own. The pattern tells you more: whether replies have been getting shorter or slower over weeks, and whether they ever pick the thread back up. Decode reads that rhythm across the whole conversation and gives you a plain read on it, with a free account.

Busy people come back, usually with some kind of acknowledgment ("sorry, crazy week"), and the gaps are uneven. Ghosting looks like a clean stop after a normal back-and-forth, or replies that thin out to nothing. If it's been a few days, one short, low-pressure message is fair. After that, their silence is the answer. For a second opinion, Decode looks at reply timing and effort across the thread.

A second message isn't desperate if it adds something new: a plan, a question, a real thought. It reads worse when it just repeats "hello??". If your last text asked something and got no answer, give it a day or two, then send one easy follow-up. Decode suggests a reply that fits your actual conversation, and you decide whether to send it.

Ask yourself what you want to happen after they reply. If it's closure, a friendship or getting back together, say that plainly or not at all. If it's 2am and you mostly want to feel less alone, wait until morning. Talking it through first helps: Aria is free with no account, and Rhea, part of Pro, will tell you straight whether it's a good idea.

The talking stage is early and moving: you're getting to know each other and plans are getting more specific. A situationship is when it has stalled: you act like a couple some of the time, but nobody will name it, and it's been that way for a while. If you've been "talking" for months with no plans that go anywhere, it's probably the second.

Dry texting is a pattern, not one "ok". Short replies, no questions back, and you carrying every conversation for days means something. One flat reply after a long day usually doesn't. If you can't tell, paste the thread into Decode: it reads effort and tone across the whole conversation, not just the last message.

Sometimes it's real, and often it's something going on in their life that has nothing to do with you. Look at what has actually changed: fewer replies, plans that keep falling through, a different tone. If it's real, a warm check-in ("hey, you've seemed quieter lately, everything ok?") usually works better than guessing. Decode works on friend and family chats too.

You can talk to Aria right now, at any hour, without an account: open Calypso and type whatever's on your mind. She's an AI, made to listen first and stay on your side. If you're thinking about hurting yourself or you're not safe, please reach a person: call or text 988 in the US, or your local emergency number.

We built it to be, because people bring Calypso things they haven't told anyone. No other user can see your conversations, and we treat everything you write as sensitive data, wherever you live. It's encrypted on its way to us and in our database, and screenshots you upload are deleted about a day after your decode. You don't need your real name or a phone number. We never sell your data or use it for ads. The Privacy Policy has every detail.

No. No one at Calypso reads your conversations, and no one watches them as they happen. A person on our team opens a conversation only if it's flagged, and only for one of these reasons: our automated safety system detects signs that someone may be in crisis (so we can check how Calypso responded, fix what went wrong, or review a safety restriction on an account); you ask us to, for example in a support email; someone reports it to us; or the law requires it. What we learn makes Calypso safer and never trains AI models. Because no one is watching live, this can't get you help in an emergency. If you're in danger, use the crisis lines on our Safety page.

No. We don't use your conversations to train or fine-tune AI models, ours or anyone else's, and we haven't opted in to any AI provider doing so. Our main AI provider's terms forbid it from training on what we send. One provider we use for some features doesn't promise in its published terms that it won't train on what it receives, and it keeps data under its own retention rules; section 7.1 of the Privacy Policy explains. If our practice ever changes, it will be off by default and opt-in only. The usage data we use to improve Calypso, such as which features get used and error rates, never includes the text of your conversations.

No, never, to anyone. We don't sell personal information, we don't share it for advertising, and we haven't done either. Calypso shows no ads. If your browser sends a Global Privacy Control signal, we honour it as an opt-out of sale and sharing.

Your data is encrypted on its way to us (TLS 1.2 or higher) and in our database, which holds your account, conversations and decodes. You sign in with a 6-digit code sent to your email; it expires after 15 minutes and stops working after 5 wrong tries, so keeping your email account secure keeps Calypso secure too. If you set a password, we store only a salted hash of it. Only the people on our team who need it to run Calypso can access personal data. No system is perfectly secure: if a breach ever puts you at risk, we'll tell you and the authorities within the time the law requires. To report a security problem, email security@trycalypso.ai.

Screenshot images are deleted from our servers about a day after your decode finishes. Chat exports and pasted text, the text we read from screenshots, the details you add about the other person and the report itself stay on your account so you can come back to the read, until you delete the decode, which removes all of it. Nothing you upload is shown to other users unless you turn on sharing for that read. If you open a chat from a read, the voice you talk to can see that uploaded conversation so it can answer questions about it; your other chats only see one line per recent read (who it was with and its headline).

A way to talk without keeping anything. Incognito messages are never saved to your history or added to Calypso's memory of you, and birth details you mention aren't stored. Dictation is switched off too. When you leave the chat, it's gone from Calypso. Each message still goes to our AI providers so they can write the reply, and they handle it under their own retention rules. Incognito needs a free account.

They're saved under a random ID kept in your browser, not under a name or email. We also use your IP address to apply the free daily message limit. Both stay until you ask us to delete them: email privacy@trycalypso.ai and tell us roughly when you chatted, so we can find them.

Yes. You can delete any conversation, any decode and any birth details you've saved yourself, at any time, in the app. To clear what Calypso remembers about you, open What Calypso remembers on your profile and forget things one at a time, or choose Forget everything. To delete your whole account, choose Delete account on your profile: our team carries it out within 30 days and emails you when it's done. Ask privacy@trycalypso.ai for a copy of your data first if you'd like one. A few records outlast the deletion for a limited time, and we'd rather you hear it from us: database backups roll off within 90 days; our analytics provider keeps usage events (never message text) for up to 12 months; our payment processor keeps billing records as long as financial law requires; we keep support emails and deletion requests for two years; and we keep safety records for up to two years after the last related event.

Yes. There's nothing strange or "less valid" about putting feelings into words with an AI listening — what matters is whether you feel a little less stuck after. Calypso was built specifically as a place to vent without judgment, at the hours when reaching out to a person isn't possible or appropriate. That said, an AI is not a substitute for a person you love or a professional you trust. If you have those, Calypso isn't here to replace them — it's the in-between, the 3am quiet space, the moment after the conversation where you need to talk it through before you can sleep.

Anything that's keeping you up. Common ones: a conversation you can't stop replaying; a friendship that's gone weird; the version of yourself you saw in a fight; whether you're overthinking; what you actually wanted to say; the question you can't ask out loud. Calypso doesn't require you to know what's wrong before you start — "I don't know, something just feels off" is a perfectly real beginning.

Yes, when you're signed in. Your conversations are saved so you can pick up where you left off, and Calypso keeps short notes on what matters to you (the people in your life, what's going on and coming up, things you like, good news you shared, your jokes) so you don't have to explain everything again in a new chat. Those notes are private to your account and are never made from incognito conversations. You can see all of them, and forget any of them, on What Calypso remembers in your profile. Without an account, chats are saved under a random ID kept in your browser, and no memory is built from them. If you want a conversation that isn't saved at all, use incognito mode.

Yes. Open your profile and choose What Calypso remembers. It shows everything Calypso keeps about you: a short note on who you are, which you can edit or add to; the people in your life; what's going on and coming up; what you like; good news you shared; your jokes; what each persona has told you about themselves; and the conversations it remembers. Sensitive things, like your health, are marked, and Calypso won't bring those up first. Memory needs a free account.

There are a few ways, depending on how much you want gone. On What Calypso remembers, tap Forget next to anything and Calypso removes it and stops bringing it up from past chats. If you change your mind, it's listed under Things you asked Calypso to forget, with Allow again. In a chat, tap one of your own messages and choose Forget this: the message stays on screen, but Calypso stops using it, in that chat and in future ones. Forget this conversation removes a chat's summary and notes but keeps the chat in your list. Forget everything clears the whole page at once and can't be undone; your chats stay, and Calypso starts keeping notes again as you keep talking. Deleting a chat removes it with its summary and notes, but things Calypso already noted from it, like a name, stay until you forget them. One thing we'd rather you hear from us: forgetting a message doesn't hide it from Calypso's safety checks in that chat. To keep a chat out of memory from the start, use incognito.

Yes. Each thread is its own conversation with its own history. You can have a vent thread, a "what did they mean" thread, and a decode thread running in parallel and they won't bleed into each other.

That's fine. Calypso's voices start the conversation — they ask what's on your mind, and they're built to be patient with "I don't know." You can describe a feeling, drop a screenshot with no context, or just say "weird day." The voice will work with whatever you give it.

Drop a screenshot, paste a conversation, upload a text file, or send a WhatsApp / iMessage export. Calypso reads tone, attachment patterns (how each person tends to show up — anxious, avoidant, secure), response time and rhythm, who puts in the effort, the small word choices that change a message's temperature, and — if you're on Pro and have entered both people's birth info — a synastry chart reading on top of all of it. The decode works on any conversation: a partner, a date, a friend, a family member, a coworker.

Yes. WhatsApp lets you export any chat from inside the conversation (open the chat's menu, then Export chat, Without media). Drop that file into Calypso and it'll parse the whole history, not just a snippet. This is the most powerful mode of the decoder because it has weeks or months of context instead of a single screenshot. The same works for iMessage exports.

The most useful screenshots include both sides of the conversation, show timestamps if possible, and capture a full back-and-forth (3–5 messages per side is plenty). You can attach multiple screenshots in a row for longer threads. Crop out anything you don't want included — Calypso reads what you give it.

No. Calypso never contacts anyone in the conversations you upload, and no other user can see your decodes. A read becomes public only if you turn on sharing for it yourself. A shared read leaves out the name you gave the other person, your screenshots and your relationship details, though it does include short quotes from the conversation, which can contain anything written in them. You can turn sharing off at any time, and the link stops working until you share it again. Uploads contain other people's messages, so please don't upload a conversation you got without permission, such as from someone else's phone.

Four parts. (1) The answer: what Calypso thinks is going on, in plain words, with any green flags and anything worth taking seriously right under it. (2) The proof: the numbers behind the read, such as who carries the conversation, how fast each of you replies and who reaches for what. (3) The depth: how each of you tends to show up (anxious, avoidant, secure or disorganised), the thread again with what Calypso noticed in the margins, and, on Pro with both birth charts saved, the chart read. (4) What now: one message you could send, with the approach behind it (hold ground, open the conversation, or hold the silence). Apart from the chart read's compatibility ratings, nothing in the report is a score, and how sure Calypso is gets said in words.

Usually under a minute. Long WhatsApp exports take a little longer because there's more to read.

Calypso reads patterns in language — the same kinds of patterns a thoughtful friend would notice if you handed them the screenshot. It does that well, and the report tells you how sure it is in plain words (clear read, fairly sure, it's close), so you can tell a strong signal from an ambiguous conversation. But the decode is one read, not the truth. The actual person on the other side of the conversation is the only one who knows for sure. Treat it like a second opinion from someone who's read a lot of conversations — useful, sometimes clarifying, never the final word.

Yes, with a limit worth knowing. A decode is a read of two people, so in a group chat Calypso reads you and the person who posts the most, sets the others aside, and says so on the report. It also holds its read a little more loosely, because a group changes how people talk. For a read of one friend, a one-to-one chat with them works best.

Not yet. Calypso reads and chats in English for now. If a conversation is mostly in another language, the decode stops and doesn't use your free read. A few words or phrases in another language are fine. More languages are coming later.

It will offer an option, not an order. The "what now" part of every decode suggests one message you could send and names the approach behind it — hold ground, open the conversation, or hold the silence — so you can see what it would mean in context. You can send it, rewrite it, or ignore it, and holding the silence is sometimes the suggestion. The whole product is built on the belief that you know your situation; what's hard is reading what's in front of you. Calypso helps with the reading.

Four, each with its own personality. Aria is the wise best friend who is on your side. Markus is the guy's view: an honest male perspective with no bro talk. Rhea is the big sister who tells you the truth and what to do about it. Luna is the astrologer, who reads your chart and theirs to make sense of what's going on. The same situation reads differently depending on who you talk to, because each one notices different things.

Aria is free, and you can talk to her without an account. Markus, Rhea and Luna are part of Pro. Inside the app you'll always see clearly what your plan includes.

Each conversation stays with the persona you started it with. To talk to someone else, start a new conversation and pick them. If you're signed in, Calypso remembers what you've told it across conversations, so you don't have to start from zero.

Start a new conversation with a different persona. Aria is free; Markus, Rhea and Luna are part of Pro.

Each persona has a distinct personality and reading style, not just a different greeting. They surface different things from the same conversation: Aria stays with how you feel, Markus reads the other person, Rhea names the pattern and the plan, and Luna reads it through your chart and theirs.

No. Calypso is emotional-support technology — a confidante you can vent to and a decoder for conversations you can't stop reading. It is not a therapist, not a counsellor, not medical care, and not a substitute for care from a licensed professional. Calypso doesn't diagnose or treat anything. If what you're carrying calls for professional care, a licensed human is the right place for it — Calypso is for the in-between moments: the 3am vent, the message you've re-read eleven times, the thought you need to say out loud before you can sleep.

For most people, in most moments — yes. The risks to know about: an AI does not have the clinical judgment to recognize a crisis or escalate it the way a trained professional would, even though Calypso has safety guardrails that flag crisis language and surface the Safety page. An AI also can't follow up with you days later to see how you're doing. If your distress is acute (suicidal thoughts, self-harm urges, danger to others), Calypso is not the right tool — please use the crisis resources on our Safety page directly.

Calypso can be a place to put anxious thoughts into words — naming what's spiralling, having something to do with the worry at 2am, getting a second read on whether what you're afraid of is what's actually happening. That can help in the moment. But ongoing anxiety, especially if it's affecting how you eat, sleep, work, or relate, deserves more than what an AI can offer. Please see a licensed mental-health professional for clinical anxiety care — and use Calypso alongside, not instead.

Plenty of users do. Things that come up between appointments — replaying a conversation, processing what came up in a session, thinking out loud before you can sleep — are a natural fit for Calypso. Calypso is not connected to anyone providing your care, doesn't share your data with them, and isn't a clinical extension of their work. It's a way to keep thinking out loud between Tuesdays.

Stop here. If you are in immediate danger or thinking about harming yourself, please reach out to a crisis line now — 988 in the US and Canada (call or text), or your local emergency number. The full list of resources is on our Safety page. Calypso is not an emergency service.

Replika and Calypso solve different problems. Replika is an AI companion you build a persistent character relationship with — over time, your Replika has memory of you, a personality you've shaped, sometimes a romantic dynamic. Calypso is a different kind of companion: several voices plus Decode, rather than one bonded character. It's built around reading conversations (the Decode pillar) and around having more than one voice to talk to depending on what you need that day. If you want a long-term AI relationship, Replika is built for that. If you want to drop a screenshot and ask "what did they mean by that," and have a confidante to talk it through with, Calypso is built for that.

Pi is a single empathetic AI voice for general conversation. Calypso has multiple voices with distinct personalities (so the conversation feels different depending on who you pick), and Calypso has the Decode feature, which Pi doesn't — you can hand Calypso a real screenshot and get an editorial read on it. Pi is excellent for general venting; Calypso is built for the specific case of decoding a real conversation and for the case of wanting different "energies" on different days.

ChatGPT is a general-purpose assistant — it can write code, summarize documents, plan a trip, and yes, also have a vent-style conversation. Calypso is purpose-built for a narrower thing: reading between the lines of real conversations and being a confidante. The voices, the four-part decode report (the answer, the proof, the depth, what now), the synastry layer, and the tone of the product are all tuned for that. If you've vented to ChatGPT and felt it was a little too eager to fix or solve, that's the difference Calypso is trying to make.

Co-Star and The Pattern give you horoscopes and personality-style readings; their compatibility features tell you about you-and-them based on natal data. Calypso's synastry feature (Pro) does that inside the context of your actual conversation — it reads your texts and overlays the chart, instead of giving you a chart in the abstract. So Calypso is less "your year ahead" and more "the argument you had on Tuesday", read through both your charts.

There's a free tier: Aria, a daily allowance of messages, and one decode every 7 days with a free account. Pro is a paid subscription, monthly or yearly, that adds Markus, Rhea and Luna, unlimited chats and unlimited decodes, and turns on the synastry read inside each decode. The current price is shown on the subscribe page before you pay, and nothing is charged until you confirm at checkout.

Not right now. The free tier is its own thing, not a trial: it never expires, and you can use it as long as you like before deciding whether Pro is worth it.

Yes. Cancel from your profile, which opens our payment processor's billing page, or email hello@trycalypso.ai. Cancellation takes effect at the end of your current billing cycle; you keep Pro access until then. We don't add cancellation friction.

No. When you subscribe to Pro, you enter your card on our payment processor's own secure checkout page. We never see or store your full card number. We keep only what we need to manage your subscription: your plan, its status and a record of each payment.

No — your conversation history is yours. You'll lose access to the voices and synastry that are Pro-only, but the threads you had on those voices stay readable from your account.

No. Calypso talks to our servers to do its reading and respond. It needs an internet connection.

Email us at support@trycalypso.ai and tell us what happened — paste the conversation if you can. We review every report and tune accordingly. If a response felt clinical or judgmental in a way Calypso shouldn't, that's something we want to know about; it's a real bug, not a small one.

Email us at support@trycalypso.ai, and — if it's urgent — reach out using the Safety page resources. We take responses on sensitive topics seriously and review reported messages quickly.

Email support@trycalypso.ai. We're a small team, and a real person reads every message. For your data (a copy, a correction or a deletion), write to privacy@trycalypso.ai; to report a security problem, security@trycalypso.ai; for a safety concern, safety@trycalypso.ai. For urgent matters, such as billing problems or account access, put "urgent" in the subject line and we'll prioritize it.

You don't need it to get in. On the sign-in page, choose Sign in with email and we'll send a 6-digit code from hello@trycalypso.ai (check spam if it doesn't arrive). The code works for 15 minutes. Changing a password you've forgotten isn't possible in the app yet, so keep signing in with a code.

Yes. Email privacy@trycalypso.ai and we'll send you a copy of your threads and decodes in a readable format.

Email support@trycalypso.ai. If it's a bug or a bad response from a voice, paste the conversation so we can see what happened. If you're reporting misuse of someone else's private messages on Calypso (for example, someone in your life is decoding screenshots of conversations they shouldn't have access to), email safety@trycalypso.ai so it reaches the right people.