It reads the emotion behind your words — before anything leaves your phone.
Two on-device classifiers analyse every message in ~80ms. The cloud is only called when confidence is low or distress is detected. Tap a mood to watch it work.
Fast and private by default. Deep only when it counts.
On-device BERT
Two classifiers — emotion (13 classes) and risk (Normal / Anxiety / Depression / Suicidal) — run locally in ~80ms. Free, and works offline.
The gate
When on-device confidence is low, the message is sent to the cloud for a deeper reply. Crisis signals trigger immediately — no network required.
Cloud AI Model
Only 30–50% of messages reach the cloud, where the Cloud AI Model responds with CBT- & ACT-grounded warmth — and points to real human support when needed.
There's always a human in the loop.
Sikivu Coach is the everyday companion — but it never stands in for people. The moments that matter most are held by humans, every time.
When risk is elevated, Sikivu points to Befrienders Kenya and real human support — always free to reach from Sikivu, and the crisis path works even offline.
Peer Circles are guided by trained community mental-health workers — people, not algorithms, hold the space.
Every feature is tested against its clinical theory and reviewed by our team before it ever reaches you.
Why this combination works.
Sikivu isn't a guess. Each ingredient is one the research already backs — here's the evidence behind every piece, and the honest line about what we can't yet claim. Every citation links to its source.
CBT and ACT are among the most-studied therapies for anxiety and low mood. A 2024 umbrella review of 39 meta-analyses found internet-delivered CBT reliably reduces symptoms across a wide range of conditions — depression and anxiety among them — with effects that vary by condition from small to large.
Umbrella review of internet-delivered CBT · Digital Health, 2024 ↗Internet-delivered CBT isn't only a lab result. A review of guided iCBT in routine clinics found it stays effective in everyday practice — with high satisfaction, acceptable adherence, and only about 3% of people worse off.
Guided iCBT in routine care · JMIR, 2020 ↗Guided programmes consistently beat unguided ones. With human support, digital CBT shows large effects; without it, the effect shrinks to small (d≈0.24). Support is the active ingredient.
Spek et al. · Psychological Medicine, 2007 ↗In a Zimbabwe trial of brief therapy from trained lay health workers, 6-month suicidal ideation was 2.3%, versus 12.3% in usual care — the model behind our Circles.
Friendship Bench RCT (Chibanda et al.) · JAMA, 2016 ↗A pre-registered observational study with IRB review and open publication. Headline metric: cost-per-WELLBY. Measurement-based care isn't optional polish — across 29 randomized trials (15,255 people) it raised response and remission rates over usual care.
Meta-analysis of enhanced evidence-based care · Translational Psychiatry, 2021 ↗The honest line: these are findings from the published research on each ingredient — not Sikivu's own results. We're pre-launch and have measured no outcomes of our own yet. When we do, every number will carry its confidence level and its source — exactly the way these do.