Calm Companion

How this companion keeps you safe

This is our safety and crisis protocol, published here on purpose so anyone can check it.

• Crisis help comes first, always. If you mention suicide or self-harm, the companion stops and shows you real help: our Get help now page with verified crisis lines for your country, and findahelpline.com. That response is fixed and written by people, not generated by AI. Nothing you say in that moment is stored anywhere; we keep only an anonymous monthly count of how often help was shown, with no names and no content. The crisis banner also stays at the top of this page at all times.
• Always an AI, and it says so. You see a clear notice when a session starts, again every three hours if you keep talking, and any time you ask. The companion never claims to be human, never claims to have feelings, and never guilts you into staying.
• Never therapy. The companion helps with everyday stress, calm habits, sleep routines, and communication. It does not diagnose, treat, or manage any condition, it never plays the role of a clinician, and it will point you to real professionals and our Get help now page when a conversation needs more than a wellness tool.
• Adults only. The companion asks for your date of birth once and stores only the yes or no answer, never the date. If it ever learns you are under 18 it stops and points to free support for young people instead. The rest of KEEP CALM stays open to everyone.
• Memory belongs to you. Memory is off until you turn it on. When it is on, the companion keeps a short list of plain notes, like a first name or a goal, never conversations, never anything about health conditions or crises. You can read every note, delete any note, delete them all, or turn memory off at any time, and deleting your account erases everything. Notes expire on their own after about a year of inactivity.
• Conversations are not stored. Chats live in your browser for the session and are gone when you close the page. Messages are processed by Anthropic, the AI provider, as our service provider under commercial terms that do not allow training on your messages. Your words are never sold, never shared for advertising, and never used to train AI models.
• No tricks to keep you here. No streaks, no "I miss you" messages, no punishment for leaving. After a long session the companion suggests a break, and it will always encourage real-world connection over more chat time.
• Check-ins are opt-in and never a ping. If you turn check-ins on, the companion greets you with one short note about what you were working on, and only when you open this page yourself. It never emails, texts, or notifies you, and it greets you at most once every eight hours, so returning often never means more nudges. Turning memory off turns check-ins off too.
• Guided weeks are optional, written by people, and yours to stop. The companion can walk you through a short multi-day arc of small practices, like a wind-down week. Every step is written by KEEP CALM, not generated by AI, and it is one small step per day, never a binge. Starting is always your explicit choice, pausing or stopping is honored instantly, missed days are never held against you, and the only thing stored is which day you are on. Stopping erases even that.
• Posters it makes pass the same wall as everyone else's. The companion can turn a conversation into a personal KEEP CALM poster, but only after you confirm the exact phrase, and every phrase goes through the site's full moderation pipeline before anything is published. If any check cannot verify a phrase, it waits for a human reviewer instead. Your calm plan is stored as ordinary memory notes: visible, deletable, and never medical advice.

Need help right now? Get help now · Find a helpline

The companion is free and is never for sale. It is excluded from every paid perk on KEEP CALM, and its safety limits are never raised for money. Read our Free Help Guarantee.