Now available for Android

Your AI chief of staff —
living on your phone.

CueInbox reads all your notifications, triages them by priority, and drafts smart replies — powered by on-device AI. Nothing ever leaves your phone.

100% on-device AI No cloud, no account No sign-up required Free forever
Smart Queue Reminders Settings
96
times a day you check your phone. Most of it is noise.

Telegram pings. Slack threads. Gmail follow-ups. WhatsApp groups. LinkedIn messages. Every app fights for your attention, but only a few messages actually need you right now. There's no single place that answers: “What actually needs my attention?”

One smart inbox. 20 apps. Zero cloud.

CueInbox captures notifications from all your messaging and productivity apps, runs them through AI that understands urgency, context, and intent — and presents a single priority queue.

Triage

AI classifies every notification into Now, Today, Delegate, or FYI. VIP contacts always surface. Noise gets filtered.

Reply

3 contextual reply options per message using the last 5 messages of conversation. Concise, warm, or direct. Tap to send.

Remind

Snooze any thread with AI-suggested timing. Get reminded when it matters, not when it arrives.

AI that actually reasons about your messages.

Every notification passes through a two-stage pipeline — instant heuristics first, then on-device LLM for hard cases. Chain-of-thought reasoning activates when confidence drops below 70%.

📱

Notification

Text + image from any app

Heuristics

Instant, always works offline

🧠

LLM Triage

Gemma 4 E2B on-device GPU

🏆

Priority

Now · Today · Delegate · FYI

💬

Summary

Thread context, last 5 msgs

Draft Replies

3 options per thread

Chain-of-thought reasoning activates automatically when confidence < 0.7 — the model explains its reasoning before scoring. Vision analysis runs on notification images and photo previews for richer context.

Works with everything you already use.

20 apps. One queue. Just grant notification access — no logins, no API keys.

TelegramTelegram
WhatsAppWhatsApp
GmailGmail
SlackSlack
SMSSMS
InstagramInstagram
DiscordDiscord
SignalSignal
TeamsTeams
LinkedInLinkedIn
MessengerMessenger
OutlookOutlook
Google ChatG. Chat
SnapchatSnapchat
ViberViber
XX
RedditReddit
ZoomZoom
NotionNotion
GitHubGitHub

Trusted by people who hate notification overload.

1.2k+
Downloads
20
Apps integrated
0
Bytes sent to cloud
Free
Forever
★★★★★

“Finally an app that just works across all my messaging apps. The AI triage is surprisingly accurate — it correctly prioritized a critical work Slack message over a meme from a friend.”

Alex M.
Product Manager
★★★★★

“The privacy angle sold me immediately. Everything runs on-device — no API keys, no subscription, no data leaving my phone. This is how AI apps should be built.”

Sarah K.
Security Engineer
★★★★★

“I was drowning in 200+ notifications a day. After a week with CueInbox I only read what the AI surfaces as “Now”. Everything else waits. Game changer.”

David R.
Founder

Your data stays on your device. Period.

Gemma 4 E2B — Google's latest language model — runs directly on your phone's GPU. 2.4 GB downloads once, then everything works offline.

What we don't do
What we do
Send notifications to any server
Process everything on-device
Store messages in the cloud
Encrypted local database
Require an account or login
Works immediately after install
Charge a subscription
Free forever, no ads

Not another todo app. Not another email client.

Todoist Notion Superhuman Spark Mail SaneBox CueInbox
Price $7/mo$10-20/mo$33/mo$8-20/mo$7-36/mo Free
Apps covered Tasks onlyGmail onlyEmail onlyEmail onlyEmail only 20 apps
AI triage NoBusiness onlyAuto LabelsBasicLearns senders Hybrid LLM
Draft replies NoNoAuto DraftsAI draftsNo 3 per msg
Privacy CloudCloudCloudCloudCloud On-device
Works offline PartialNoNoQueuesNo Full
Cross-app NoNoNoNoNo Yes

Up and running in 3 minutes.

1

Install

Download the APK and grant notification access. No account needed.

2

Download AI

The Gemma 4 E2B model downloads once (~2.4 GB over Wi-Fi). Takes a few minutes.

3

Done

Every notification gets triaged automatically. Open the app to see your priority queue.

Actively developed. New features every week.

CueInbox ships fast. Here's what's landed recently.

v0.9.1
Apr 2026

Smart notification parsing

  • NotificationType enum for all 20 adapters — richer context per source
  • Vision + audio with graceful fallback when hardware unsupported
  • Deep links, better replies, extended thinking mode
  • Single draft reply instead of 3 drafts cluttering the UI
v0.5.0
Apr 10, 2026

Instant onboarding

  • Model download no longer blocks onboarding — land in the app instantly
  • Background download with live progress banner in Smart Queue header
  • AI features activate automatically when download completes
v0.2.0
Apr 7, 2026

20 apps, redesigned queue

  • Support for 10 more apps: Instagram, Discord, Signal, Teams, LinkedIn +
  • Queue redesign: colored bucket headers, source-app badges, timestamps
  • Reminders overdue highlight, VIP/Mute controls, draft reply cards

Common questions.

CueInbox uses a two-stage approach: lightweight heuristics run instantly on the CPU for every notification. The Gemma 4 E2B model only activates when deeper analysis is needed (typically 10–30% of notifications). In practice, battery impact is comparable to having Telegram running in the background. The model doesn't run continuously — it processes one notification at a time and then releases GPU resources.
Android 13+ (API 33) with an arm64 processor and GPU. This covers the vast majority of Android phones released after 2020, including all modern Samsung Galaxy, Pixel, OnePlus, and Xiaomi devices. The LiteRT-LM SDK requires GPU acceleration — devices without a supported GPU fall back to heuristic-only triage without the LLM features.
Yes — fully. The only time you need internet is the initial one-time model download (~2.4 GB over Wi-Fi). After that, triage, draft replies, reminders, and all AI features work completely offline. No API calls, no cloud sync, no telemetry.
CueInbox requests Android notification access — the same permission that allows a Wear OS watch to mirror your phone's notifications. It reads notification text, sender names, and optionally attached images. All of this data is processed locally and stored in an encrypted on-device database. Nothing is transmitted anywhere.
The app itself is under 30 MB. The Gemma 4 E2B model file is ~2.4 GB and downloads once on first launch over Wi-Fi. After that, no additional storage is used unless you accumulate a large notification history (which can be cleared in Settings).
No. CueInbox never sends anything without your explicit tap. Draft replies are suggestions — you review them, edit if needed, and choose to send. The app has no ability to send messages autonomously.

Take back your attention.

Download CueInbox and let AI handle the triage.

Get in touch

Have questions, feedback, or found a bug? We read every email.

info@cuelinbox.com