Presence & proactive check-ins

Most assistants only answer when asked. MIRA can start the conversation — a periodic check-in to see how you’re doing, or a morning briefing built from your calendar, wiki, and recent activity. This guide turns those on.

Both features are per-user and opt-in, and both need a connected channel so MIRA has somewhere to reach you. If you haven’t set one up yet, do that first: see Connect a channel. Telegram is the quickest to start with.

You enable and tune companion behaviour on the Presence page (rhythm, personality, what MIRA sends, and the daily briefing) — or just by asking MIRA in chat. Device pairing, browser push, and the test buttons live under My Preferences → Notifications & devices, which every user can reach. (When you finish onboarding, MIRA already configures Presence from your answers — admins are switched on automatically; others enable once a safety contact is set.)

Turn on companion check-ins

Companion mode is what lets MIRA message you on its own — gentle, periodic check-ins rather than replies to your prompts. It’s designed to feel like a friend who knows you, not an alarm clock: MIRA varies its timing, learns when you tend to reply, and backs off when you’re brief or busy.

Go to the Presence page and enable companion mode for your account.
Choose a preferred channel for the check-ins (for example Telegram or Signal). This is where the proactive messages will arrive. If you leave it unset, MIRA uses whichever surface you last talked to it on — including the web app or the native mobile app, which both receive check-ins and the morning briefing as push notifications (register a device under My Preferences → Notifications & devices so a check-in can actually reach you there).
Set your quiet hours — the windows when MIRA must never message you (for example overnight, or during a regular nap). MIRA respects these absolutely; it won’t reach out inside a quiet window no matter what.

That’s enough to start. MIRA picks varied times within your allowed windows, skips a check-in if you’ve just been chatting, and adjusts how often it reaches out based on whether you tend to engage.

Companion mode is usually set up for someone — an admin or family member configures it on the person’s behalf — but you can also enable it for yourself. Either way it’s strictly per-user: people without it enabled get normal MIRA behaviour.

Tune the rhythm (Settings → Presence)

MIRA’s timing is a fuzzy band, not a fixed schedule:

Messages per day — a range like “1–4 a day”. MIRA picks a count in that band each day and scatters them at varied, non-round times inside your contactable hours (each at least a minimum gap apart). “Message me once or twice a day.”
Minimum gap — the least time between two reach-outs (default 90 minutes).
It leans in when you’re engaging and backs off when you’re brief or not replying — and pauses after a few unanswered, resuming the moment you reply, so MIRA never talks into the void.

Prefer predictable times instead? Switch the rhythm to Scheduled and give it fixed times (e.g. 09:00 and 18:00). Everything is also adjustable in chat — “only check in in the mornings”, “leave more time between messages”.

Give MIRA personality (what it sends, and how)

On the Presence page you also shape what a reach-out is and how it sounds:

Message types — toggle which kinds MIRA may send: a check-in, a joke, “what I’ve been up to” (it’ll genuinely mention what its background agents did for you), a follow-up on something you recently discussed, a share, or a bit of encouragement. MIRA picks one per reach-out, biased by context.
Tone — three sliders (warmth / playfulness / verbosity) with quick presets (“Warm & chatty”, “Calm & concise”, “Playful”, “Professional”); for deeper voice, edit MIRA’s persona wiki page.
Or just say it: “be funnier”, “stop the jokes”, “keep it short”.

Each person’s settings override the instance-wide defaults an admin sets in the companion config block. The daily briefing is separate and isn’t affected by these limits.

The care network

Presence can be more than friendly company: it can be a quiet wellbeing net for someone you look out for — a child, or an older parent living independently. If the person seems to be having a hard time, or simply goes quiet, MIRA gives a trusted contact a gentle heads-up so a real human can check in. You set this up on the Settings → Presence page under Care network.

It’s built on four principles:

Never covert. MIRA always tells the person, in plain language, that it’s looking out for them and may give their contact a heads-up. The first time a care arrangement is active, MIRA works this disclosure naturally into a check-in. The “the person knows” acknowledgement on the Presence page records that this has happened.
Concern, not tattling. Only genuine signals escalate — clear distress, or a run of unanswered check-ins. An ordinary off day doesn’t trigger anything; MIRA just responds warmly. Repeat alerts for the same signal are suppressed.
A heads-up, not 911. The contact gets a short, factual note — “you might want to check in” — so a person who knows them can follow up. MIRA escalates to that human; it does not call emergency services. In parallel, MIRA responds to the person warmly and surfaces crisis resources to them.
Minimal disclosure. The contact sees a one- or two-sentence summary of the signal and the person’s name — never the full conversation.

Choose a care role

On the Presence page, pick who Presence is for:

Just me — a companion for yourself. No one gets an ongoing wellbeing view of you, but your safety contact is still alerted in a genuine emergency (distress or a run of unanswered check-ins) — the care network isn’t only for children and older adults, and an ordinary adult can absolutely want people to turn to when something’s wrong. This is the default.
An adult who needs support — for a grown-up who wants someone actively looking out for them: recovery, a chronic condition, living alone, a hard stretch. Like the child/older-adult roles, a care contact is alerted on distress or silence and — once you’ve been told and have consented — can see your aggregate wellbeing view. The difference is tone: it’s a peer arrangement, no age framing.
A child — a guardian is alerted if their child seems to be struggling; MIRA keeps a gentle, age-aware tone.
An older adult — a contact is alerted on silence or signs of distress — a light-touch wellbeing check for someone living more independently.

The child and older-adult roles additionally expose a quiet wellbeing view to the guardian (aggregate engagement over a trailing window — never the person’s messages), gated on the person having been told. The Just me role never exposes that view; it’s the framing/consent that differs, not whether a contact is reached in an emergency. How serious a signal is tunes the message: an acute signal (mentions of self-harm, or acute physical symptoms) sends an urgent heads-up and shows the person crisis resources prominently; a concerning one (a low mood, loneliness) sends a softer “you might want to check in”.

Set a safety contact

When you configure check-ins, set a safety contact: the person MIRA notifies if check-ins go unanswered (three in a row over 48 hours) or if a message reads as distress. (Choosing a care role above sets this same contact.) Pick them by searching for their name — you don’t need to know their internal user id.

You can name more than one contact — for example both parents of a child, or two adult children looking after an older parent. A distress or missed-check-in alert reaches all of them, and it counts as delivered if it reaches any one, so a single unreachable contact never silences the alert.

The safety contact is a hard prerequisite for ordinary user accounts — you can’t enable companion check-ins without one — because a feature that holds daily conversations with a vulnerable person needs somewhere to turn when something’s wrong.

The heads-up reaches the contact on their own messaging channel — whichever they last used with MIRA (Signal, Telegram, email, and so on) — so it lands on their phone, not just a web page they may not have open. It’s also recorded in a “Safety alerts” thread in their MIRA web account as a durable copy, and a contact who only uses the web still gets it there. If the messaging channel can’t be reached, MIRA falls back to the web record and flags the delivery problem on the health page rather than failing silently.

Turn on the daily briefing

The daily briefing is a single morning summary, assembled from:

your calendar — what’s coming up today,
recent wiki updates — notes you’ve added or changed,
recent activity — automation runs and the like,
pending approvals — anything waiting on your sign-off (see below).

To enable it:

On the Presence page, enable the daily briefing.
Set the hour it should arrive (for example 7 for 7am). You can change it any time — or just ask MIRA: “set my briefing hour to 7”.

The briefing arrives on your connected channel. If your host was asleep or restarting at the scheduled time, the briefing catches up later the same day.

Stay on top of pending approvals

Some things MIRA proposes wait for your explicit sign-off before they take effect — for example a schedule the agent created for you (when the approval gate is on) or a wiki edit sitting in the review queue. Left alone, these can quietly pile up.

So MIRA reminds you: when anything is awaiting approval, your daily briefing and your companion check-ins include a short nudge — “you have 2 schedules and 1 wiki edit waiting for your approval”. (A briefing with nothing else going on will still send when approvals are pending, so they don’t slip through.)

You can act on them right there by replying — no need to open the dashboard:

“summarise what’s pending” → MIRA lists each item (what the schedule does, what the wiki edit changes).
“approve all”, or “approve the backup schedule” → MIRA approves what you named (or everything).

The reminder is just a reminder: MIRA only approves when you tell it to — it never approves on its own, and never inside the proactive message itself. The human-in-the-loop review stays with you. (You can still approve from the Automations and Wiki review pages in the UI if you prefer.)

Test it right now

You don’t have to wait for the next scheduled moment to confirm everything works. Under My Preferences → Notifications & devices there are two test buttons:

Send a check-in now — fires a companion check-in immediately.
Send a briefing now — starts building today’s briefing and delivers it on your channel when it’s ready. Because a briefing is a full generation pass, the button returns right away and the briefing follows a few moments later (it doesn’t block while it’s written) — so it won’t hang the app on a slow model.

Use these to confirm the message actually reaches your chosen channel before relying on the schedule.

Troubleshooting

Nothing arrives. Confirm you have a connected channel and that it’s set as your preferred channel for notifications. On Telegram, MIRA can only message you proactively after you’ve linked your chat (Settings → My channels → Link Telegram, then send the LINK-XXXX-XXXX code to the bot) — that’s how it captures your chat id.
A check-in never came. Check-ins and briefings only fire while MIRA is running at the scheduled moment. If the host was asleep or restarting, that window can be missed — the briefing catches up later the same day, but check-ins don’t replay.
I want to pause for a while. Just tell MIRA — for example “pause the check-ins for the weekend”. It accepts the request without pushing back.

Connect a channel — set up where MIRA reaches you.
Voice replies & talking to MIRA — have check-ins and replies arrive as voice notes.
Channels — how MIRA’s channel model works.