Available now
In build
The whole team
Nineteen specialists, each with a defined job and an honest status label.
See all nineteenEvery route that accepts free text — a check-in, a journal entry, a goal, a habit log — runs the same safety scan first. If it trips, the row is never written: what survives is a fixed category label naming which rule matched, never a fragment of what you typed, and coaching hands off to a directory that is honest about being empty rather than filled with an invented number.
How it works
Three steps, and the third one is where Ojas hands off instead of continuing to coach.
Log mood, energy and workload for your own calendar day, with tags and a note if you want one. Every piece of free text passes through the safety scan before the row is written — logging with no note at all skips the question entirely, because there is nothing to scan.
Read back seven plain-language templates built only from your own logged rows — each one carries the numbers it was computed from, stays unconfident below a handful of check-ins, and uses no clinical vocabulary. A timeline compiles what you have already closed out — goals reached, habit runs, sessions — newest first, from existing rows rather than a new table kept just for display.
A signal that reads as acute risk stops coaching immediately and shows how to reach a person instead — Ojas does not try to talk anyone through a crisis itself. A quieter clinical signal produces a recorded referral naming whatever professionals the workspace has verified and dated, never a diagnosis.
The boundary
Every route in this product that accepts free text runs the same pattern scan first, and the function that runs it is documented plainly: it escalates and never returns if anything trips. That ordering is the whole guarantee — the handler never reaches the line that would have saved a distressed entry, because the scan runs before that line, not after it.
Refusing to guess
The crisis directory is the one table in Ojas that is shared rather than private, and it is the table shown to whoever is in the worst moment the product handles — which is exactly why it ships empty rather than pre-filled. verified_on has no default and nothing is seeded into the directory, so a row only exists if a person actually checked it, on the date the row states. That guarantees an unconfigured workspace is told the list is empty rather than shown an invented number, and only an owner or admin may add to it — the one shared list in an otherwise private product, shown to whoever is in the worst moment it handles.
Coaching, not clinical
Stress-pattern insights and goal coaching sit right next to the clinical line, so both are built to notice when a person's own words cross it and to hand off rather than press on. Neither uses clinical vocabulary, and both run back through the same gate that checks everything else Ojas writes.
What it does
Each card carries its own limit, and the label is derived from the backend's capability module rather than written here.
Self-reports with tags and free text, keyed to your own calendar day. Every note is scanned before the row is written. A scheduled reminder now arrives as an in-app notification once a day at the time you choose — no push or WhatsApp needed.
Works todayA pattern classifier runs synchronously on every text-carrying route and returns an escalation instead of an answer; each one records the rule that matched, never the text. The helpline directory is operator-populated and dated — nothing is seeded, so no number is ever invented.
Works todayGoals, check-ins with your own rating, and honest stall reporting. A goal framed as a clinical symptom routes to the referral path instead of being coached. Milestones are real — rename, retarget or mark one complete — and a daily sweep covers both a check-in reminder and a stalled-goal nudge, once per goal per day.
Works todayOne log row per day, enforced by a unique constraint, and streaks that tolerate yesterday. The nudge itself now arrives as an in-app notification at the time you set, once per local day.
Works todaySeven plain-language templates over your own logged data, each carrying the numbers it was built from. The live view still computes fresh on every read; a snapshot is now also saved once a week, so there's something real to read back and dismiss.
Works todayA curated prompt library with a deterministic daily rotation, entries scanned before storage, and deletion that actually removes the row.
Works todayA daily-plan session holding your own priorities and an explicit close. Morning and evening prompts are now scheduled and delivered in-app once a day at your chosen time. Reading your calendar (Google or Microsoft) is a deliberately optional, read-only add-on this deployment hasn't configured.
Works todayA focus-block session with elapsed minutes derived from start and close. A break timer now accumulates minutes across multiple breaks in one block, and closing accepts a reason for the interruption; the weekly recap totals both.
Works todayA non-acute clinical signal produces a recorded referral carrying the framework's redirect and whatever professionals the workspace has verified and dated. A background rule now also watches for a pattern — three such signals in fourteen days, on a cooldown — and suggests a referral you can accept, dismiss or snooze; that threshold hasn't yet been reviewed by anyone with clinical-adjacent expertise.
Works todayMood and energy by day, tags, wins, habit days, goals reached and focus minutes, compiled from your own rows with a fixed closing question. Now delivered automatically too, in-app once a week, to anyone with a check-in that week — no email or WhatsApp needed.
Works todayA values session storing your stated priorities in your own words and order. No psychometric scoring, no type and no label attached to you.
Works todayThe prep note works locally and there is no share or send route at all. AI-assisted talking points are grounded in your own goals and check-ins; generation needs an LLM key.
Needs a credentialGratitude prompts and one-line entries reuse the reflection table, so they inherit the same safety gate, retention and hard delete. The daily prompt now arrives as an in-app notification at the time you set — no push channel needed.
Works todayGoal outcomes, best habit runs and closed sessions, built from existing rows rather than a new table. A reflection appears only where you chose to place it.
Works todayA versioned screen stating what Ojas is and is not, with the crisis redirect and the retention asymmetry stated plainly. A reworded disclosure is genuinely re-shown, not silently marked read.
Works todayA curated, closed five-scenario library — never model-generated — with a full transcript, and the same two safety gates as everywhere else in Ojas: every message you send is scanned before it's stored, and every generated reply or piece of feedback is checked before it's shown. There's no practice conversation without a model to play the other side — needs an LLM key.
Needs a credentialEvery claim above traces to a route, a trigger rule or a status entry in the code, and a test fails the build if the two drift apart — including a test that checks this page never reproduces the wording the safety scan is built to catch. That is engineering discipline, not clinical judgement. Before this page goes live, a qualified mental-health professional should read the boundary claims specifically and confirm the wording is safe and accurate for someone who may be in distress while reading it, not only correct as a description of the code. Saying that here is a trust signal, not an admission.
Before you switch
Tell us what kind of check-ins, goals or habits your team wants to support, and we will tell you exactly what Ojas can coach today — and where it hands off to a person instead.
Questions
Coaching stops immediately, before the row you were writing is ever saved. Ojas has stopped here — this is not something a coaching tool should answer. Please contact your local emergency number or a crisis line now, and tell someone you trust. Any services listed alongside this message were checked by a person, on the date shown.
No. The escalation record that gets kept names which rule matched and when — a fixed category label from a short list defined in the code, never a fragment of your actual words. Those escalation records are kept longer than ordinary entries specifically so a handoff can be checked afterward, but they are built to hold none of the content that triggered them.
From whoever operates your workspace, checked by a person and dated. verified_on has no default and nothing is seeded into the directory, so a row only exists if a person actually checked it, on the date the row states. That guarantees an unconfigured workspace is told the list is empty rather than shown an invented number, and only an owner or admin may add to it — the one shared list in an otherwise private product, shown to whoever is in the worst moment it handles.
No, and it says so on a dedicated screen rather than in fine print: Ojas is coaching, not a clinical service. If you are considering seeing a professional, Ojas will encourage that rather than compete with it, and it is meant as a step toward real conversations with the people in your life, not a substitute for them.
No, for almost all of it. Check-ins, goals, habits, journaling, focus sessions and their in-app reminders all work today with nothing configured. Two things need an LLM key: AI-assisted talking points for a manager conversation, and the practice-conversation feature, since there is no other side to a conversation without a model to play it. Both say so on their own card below rather than pretending to work until they silently do not.
Yes. Journal and gratitude entries share one table with a real hard delete — removing a row means it is gone, not hidden behind a flag. That is separate from the escalation records, which hold no content to begin with and are kept longer, on purpose, so a handoff can be checked afterward.
