for therapists and counselors

your client's hardest hours are usually not in your office.

sanctuary is built for the 3am between sessions. This page is written for a clinician deciding whether to mention it to a client. It says plainly what sanctuary is, what it is not, and where its limits sit, because if you are going to lend a client your credibility, you deserve the fine print first.

what sanctuary is

sanctuary is a free set of somatic and expressive practices for heartbreak, grief, and anxious attachment. It runs in the browser. There is nothing to install, no account is needed to start, and what a person writes stays on their own device.

The practices are evidence-informed. Each one is drawn from a research tradition we can name: expressive writing, urge-surfing, slow-exhale breathing (the tradition behind HRV biofeedback), sensory grounding, behavioral activation, attachment theory. We keep the full mapping, practice by practice and researcher by researcher, at /evidence. When we say evidence-informed we mean the underlying technique has a research lineage, not that sanctuary's version of it has been through a trial. sanctuary applies the techniques; sanctuary itself has not been studied.

what it is not

sanctuary is not therapy, not a medical device, and not crisis care.
It does not diagnose. It does not treat. It does not track symptoms, score mood, or generate clinical data of any kind. And it does not replace you. It is a set of practices for the hours when your office is closed and nothing else is open, built to sit alongside clinical work, never in front of it.

You will not find treatment claims or proof language anywhere on this site, because neither would be true of sanctuary, and we would rather earn your referral with accuracy than with adjectives.

why it will not undermine your work

The most common clinical objection to apps in this space is not that they are useless. It is that they are engineered for retention, and retention mechanics work directly against therapeutic goals. sanctuary is built on a design commitment we call the First Law: absence never costs anything. In practice that means:

Two more commitments matter clinically. First, crisis language always routes to the 988 Suicide and Crisis Lifeline, and that routing is built into the page itself: it never waits on the network, so it works even when a connection is bad or a server is down. Second, dove, the AI presence inside sanctuary, is clearly labeled as AI, never claims to be a person, and after a handful of exchanges points people back toward real humans: a friend, family, a therapist. If sanctuary ever competes with therapy in a client's mind, we consider that a failure of ours.

what clients can use between sessions

Mapped to the need you will actually hear about in session. Each card names the research tradition the practice draws on; the tradition studied the technique, not sanctuary.

acute activation
the ground
For the client whose body is spiking and whose skills are not landing yet. A double inhale and long slow exhale, drawn from the cyclic sighing work of Balban and colleagues at Stanford, plus five-senses grounding as taught in trauma-informed care. At /ground.
contact urges
protest interrupt, and did not send
For the 11pm urge to text, call, or check the profile. Built on urge-surfing from Alan Marlatt's mindfulness-based relapse prevention: the urge is treated as a wave that crests and passes rather than a command. Applying it to contact urges after a breakup is sanctuary's adaptation, not something Marlatt's trials tested. At /interrupt.
expressive writing
the unsent letter
Everything the client would say to the person, written in full, and the letter goes nowhere by design. This is James Pennebaker's expressive writing paradigm; the literature reports modest, repeatable effects, and we say modest rather than inflating them. At the letter you never send.
3am loneliness
the window, and the grove
Quiet presence for the hour when no one is awake: a real night sky, and a space where others are silently alongside. The framing draws on co-regulation research, including James Coan's social baseline work on how the presence of another person changes threat processing. That research is about human presence; whether an honestly labeled digital presence carries any of the same weight is an open question, and we treat it as one. At /alone and /grove.
attachment moves
the walks
Guided sequences for the first night, the empty sunday, the hollow. They use four words for what a nervous system does after rupture: protest, withdraw, repair, freeze. The frame comes from attachment theory, from Bowlby's work on separation and protest through Ainsworth and the adult attachment literature. It names the move a client is in and deliberately never types the person. At /demo.

privacy answers you will be asked

Clients ask therapists about privacy before they ask us. Here is what to tell them, precisely:

notes stay on device
What a client writes in the practices is stored in their browser, on their device. It is not uploaded to our servers, so there is no server-side copy of their writing for anyone to request.
gift links carry content only in the fragment
When someone shares a practice with a friend, the content travels in the URL fragment, the part after the # that browsers do not send to servers. The link works; we never see what it carries.
no ad targeting of content
The content of what a person writes or does inside the practices is never used for advertising or ad targeting. Grief is not an audience segment.

The full privacy architecture, including where these guarantees end, is documented at /trust and /safety.

how to refer

1
It is free. There is no cost conversation to have, no trial that expires, and no card involved.
2
Give the client one address: joinsanctuary.io/tonight. It meets them where tonight actually is and routes them to the right practice from there.
3
There is nothing to set up. No code, no account, no onboarding call. It works the first time, on a phone, at 3am.
4
If you want to vet it first, /evidence lists the research tradition behind each practice, named honestly, including where the evidence is thin.

honest limits

sanctuary as a product has not been studied. The techniques inside it have research lineages; the product does not, and until it does we will keep saying so on every page a clinician might read.

It is not suitable as a sole support for severe presentations: active suicidality, psychosis, severe eating disorders, acute trauma responses, or anything else where a self-directed browser tool is obviously not the right level of care. For those clients, sanctuary is at most a small adjunct between the supports that matter, and possibly not even that.

You know your client. If your clinical judgment says a tool like this would feed avoidance, fuel rumination, or substitute for the harder work, trust that judgment and skip us. This page exists so you can make that call with accurate information rather than marketing.

if a client is in crisis
988 suicide & crisis lifeline
Call or text 988. Available 24/7 in the US. Free, confidential support from trained human counselors.
crisis text line
Text HOME to 741741. Trained crisis counselors, any time.
international
findahelpline.com: crisis lines in your country.
sanctuary cannot help in a crisis. When crisis language appears inside the product, the ordinary experience suspends and these resources surface immediately, without waiting on a network. Your crisis protocols always come first.

questions clinicians ask

what does sanctuary cost a client?
Nothing. The practices are free, there is no trial clock on them, and no card or account is needed to start. A paid tier exists for people who want more, but nothing a clinician would refer a client to sits behind it.
does sanctuary share data with me as the clinician?
No, and there is mostly nothing to share. A client's notes and writing stay on their device. There is no clinician dashboard, no progress report, and no symptom tracking. If a client wants to bring something they wrote into session, they bring it themselves.
what about clients with suicidal ideation?
sanctuary is not crisis care and should never be part of a safety plan as a crisis resource. When crisis language appears, sanctuary suspends the ordinary experience and routes to the 988 Suicide and Crisis Lifeline, and that routing is built into the page itself rather than waiting on a network call. For clients at acute risk, your clinical protocols apply, not a website.
can clients become dependent on it?
It is built to be outgrown. There are no streaks, no scores of the person, no engagement notifications, and nothing that expires or punishes absence, so there is no mechanic for dependence to attach to. The AI presence, dove, is clearly labeled as AI, never claims to be a person, and points people back toward humans.
has sanctuary itself been studied?
No. The practices are evidence-informed, meaning each one draws on a named research tradition such as Pennebaker's expressive writing or Marlatt's urge-surfing, but sanctuary as a product has not been independently studied and we do not claim otherwise. The full mapping of practice to tradition is at /evidence.