Running a Session

A session is where the deeper work happens. This guide walks you through starting, navigating, and finishing a session, as well as reviewing past sessions.

Starting a Session

Tap sit with… on the Dashboard, or go to Session > New. A brief sit-down screen appears — take a breath and tap begin to open the session. The default type is Free Dialogue, an open conversation with your inner parts where you let whatever needs to emerge come forward. Starting from the Dashboard (or from a part's profile) focuses the session on that part; from Session > New with no part chosen, it opens with your most recently active part.

To start a different session type, go to the Templates page and use — or create — a template set to that type. The Templates page is the only place that lets you choose the type; the five types are:

Free Dialogue — open conversation with your inner parts; let whatever needs to emerge come forward
Part Exploration — get to know a specific part more deeply; understand its role, fears, and needs
Inner Conflict — work with two or more parts that are in tension with each other
Guided Exercise — follow a structured exercise for grounding, self-compassion, or integration
Journal — free-write about your inner world; the AI helps identify which parts are present

You choose a mood when you close the session, not before — see Ending a Session below. That leaving mood feeds your mood trend data in Insights.

The AI consent screen (first AI interaction)

The very first time you open a sitting — or the first time you save a journal entry with AI help — a screen headed your choice about AI, with the eyebrow before your first sitting, appears (full-screen before a sitting; a centred overlay in the journal). It explains that this is a space for self-reflection with an AI companion (not therapy and not a crisis service), that the companion can get things wrong, and that you are always in charge. It also names where your words go to power the conversation: what you write is processed by Anthropic (and sometimes OpenAI); anything you speak is sent to OpenAI's Whisper to be transcribed; and spoken replies use ElevenLabs. None of it is used to train those models. It closes with a crisis line (Samaritans, the 988 Lifeline, and befrienders.org) and a link to the full privacy policy.

Tap i understand — let's begin to proceed. Your acknowledgement is remembered — the screen reappears if you withdraw AI consent in Settings and then use an AI feature, so you can choose again. In a sitting, choosing back to my space returns you to your Dashboard without opening the session; in the journal, the equivalent choice is save without AI, which keeps the entry but skips the AI reflection. It is worth knowing this screen exists so it isn't a surprise the first time you sit, or when you help a participant do so.

During the Session

Once the session begins, you will see a conversation interface:

Type or speak your messages in the input field at the bottom. The AI facilitator responds in character, guiding the dialogue.
Part selector — use this to switch which part you are speaking as. This helps the AI track perspective shifts in the conversation.
Voice input — tap the microphone button to speak instead of type. In modern browsers, tapping the mic records a short clip; tap it again to stop, and the clip is sent to OpenAI's Whisper service to be transcribed (see the privacy policy for details). While it processes, the button shows a transcribing… state, and once it finishes the transcript is added to the input field for you to review and edit before sending. Where your browser can't record in this way, we fall back to live transcription that appears as you speak. (Requires microphone permission.)
Text-to-speech (TTS) — if enabled in your Settings, the AI's responses are read aloud to you via the voice you have chosen. A small speaker icon on each facilitator message lets you replay individual lines.
Save a noticing — long-press any message and choose "save as a noticing" to flag it as something worth keeping. Noticings are collected on your Insights page.

Every message you send is quietly screened for safety signals before the facilitator replies. If something you share reads as tender but not urgent, a gentle caution note appears above the input — a quiet invitation to slow down, with no resources and no alarm, dismissible in a tap. If something reads as a moment of real crisis, a crisis card appears instead: it names real, reachable helplines for your region and links through to crisis lines by country. The card sits above the dialogue so it can't be missed, but it never closes the session or blocks you from continuing.

The full safety model — how the two layers work, what triggers each response, and why sittings are bounded — is set out at /safety. The design pattern behind bounded depth-work sessions is published openly at /pattern, and the evidence question behind the app is pre-registered at /evidence.

Session Length

Sessions have a natural arc rather than an open-ended run. After an extended exchange the facilitator will begin guiding the conversation gently toward closure — inviting a final noticing, a moment of acknowledgement, or a soft landing. Once the sitting reaches its end point, the session closes with a warm closing message and the end-of-session ceremony opens automatically.

In some circumstances the container closes the sitting rather than waiting for you to tap end. When this happens, the closing ceremony says so plainly — you will see a note explaining that the sitting closed itself, with a link to /safety where the reasoning is set out. This is not a limit you hit; it is a shape the sitting was given. Whatever you did not reach will keep, and you can begin a new sitting whenever you are ready.

If you are in distress at any point, the facilitator will always prioritise your wellbeing over the length arc, and the session will not close while a grounding response is needed.

Ending a Session

Tap end this sitting when you are ready to close. A closing ceremony appears where you can:

Choose your post-session mood from the mood vocabulary
Optionally write a noticing to keep — one thing you want to carry forward from this sit

Session Summary

After ending a session, you are taken to the Session Summary page (/session/[id]/summary). This is a quiet ceremony page that shows:

The part and Self side-by-side — a visual reminder of who was present in the session
A tender headline — a softened reflection of how the session moved, drawn from your mood shift
The last thing the facilitator said and the last thing you said — anchors to close the conversation
A noticing to keep — a key insight or observation from the session to carry forward
Your leaving mood — the post-session mood you rated as you closed

From the summary page you can also add a note — a brief reflection you want to attach to this session for future reference.

The summary is accessible at any time from the session detail page or your History list.

Sharing a Session

On the session summary page, tap share this with someone you trust to open the share panel. From there you can create a private link and send it to a therapist, coach, or close friend.

What the link includes:

The written reflection from the sitting (if one was saved)
Any noticings you marked during the session

What the link never includes:

The conversation itself — the full transcript is never shared

Each link is unguessable and expires automatically after 30 days. You can revoke any link at any time from the same panel on the summary page; once revoked, the recipient sees a "link not available" message. You can have up to five active links per session.

Your Session History

The History page (/history) is a paginated list of all your sessions, newest first. Each entry shows:

Session type badge — a small icon indicating the type (Free Dialogue, Part Exploration, Inner Conflict, Guided Exercise, or Journal)
Title and date — the session title (e.g. "Sitting with The Critic") and when it took place
Message and insight counts — how many messages the session contained and how many noticings were saved
Status — whether the session is active or closed

Tap any entry to open the session detail page, where you can read back through the conversation as it unfolded. The list loads 50 sessions at a time; scroll to the bottom and tap Load more to see earlier sessions.

Revisiting a Past Session

To revisit a past session in full, open it from History. A completed sitting opens in read-only form, so you can move back through the conversation exactly as it happened.

Replaying a Session

To hear a sitting again, open its summary page and tap listen to this sitting again. This opens the replay page (/session/[id]/replay). Only your own sittings open there. It is a recording rather than a dialogue: nothing you do on this page adds to the conversation.

When the page opens, no lines are showing — it says press play to hear this sitting again. Lines then appear one at a time, each in a bubble in the style of the live sitting. The facilitator's lines are labelled they say. A line you sent while speaking as one of your parts is stored as that part's line, and the replay shows it in the same style but labelled with the part's name (or, if the part has no name, its role, such as the inner critic); the live sitting always says they say. Your other lines are labelled you, from self. The line being played is full strength and earlier lines fade slightly. A sitting with no messages says this sitting has no words to hear yet, and play is disabled.

The controls:

Play / pause — the large round button in the bar at the bottom. Pausing stops playback, including any voice mid-line. Playing again starts the current line over from its beginning, and carries on from there. If the current line is the last one, playing starts again from the first line.
Previous line / next line — the skip buttons either side of play. Either one pauses playback and moves one line; tap play to play from that line. If you skip to the last line, play starts the sitting again from the first line instead.
Playback speed — the selector at the right of the bar, from 0.5× to 1.5× (0.5×, 0.75×, 1×, 1.25×, 1.5×).
Mute / unmute — the speaker icon at the top right. Whether it starts on follows the let parts speak aloud toggle in Settings. The mode for a run of playback is settled when you press play. With the speaker on, each line is read aloud in full before the next begins; with it off, the replay is silent and moves on after a pause based on the length of each line. Tapping the speaker during playback does not switch that run over: the voice carries on (or the silence does) until you pause and press play again, and only then does the new setting apply. The caption described next does follow the icon straight away, so it can appear or disappear while the sound is still in the earlier mode. When the current line has a part attached (including a line you sent as that part) and that part has a voice chosen, a small caption under the header names the part and the narrator voice reading it, for example Maren in Elara, where Maren is the part's name (the narrators are Elara, Sienna, Iona, Arjun, Silas and Marcus, depending on each part's voice). A part with no name is captioned by its role instead, such as the inner critic in Elara. Lines with no part attached, such as the facilitator's, are read in the replay's fallback voice, Sienna, and have no caption. When the speaker is off, the wait per line is 40 milliseconds per character divided by the speed, but never less than a second and a half, so a short line waits a second and a half at every speed.
Back arrow — the top left; returns to the session summary.
Progress line — the thin line under the header shows how far through the sitting you are.

The ambient sound player is available on this page too. If reduce motion is on in Settings, the page scrolls to each new line without the smooth animation.