Features
The features other auto-reply apps don't have.
Persistent memory. Your business knowledge. Cross-channel context. Real voice training. All running on your Android phone.
01
Persistent memory
Most apps forget the conversation after you close the chat. WhatsAssist doesn't.
Every message you send and receive across every connected channel is stored in a personal memory — encrypted, on your device first, synced to the cloud only if you enable backup. When someone messages you, WhatsAssist surfaces what matters:
- What this person asked or ordered before
- Promises you made ("I'll send the link tomorrow")
- Personal details worth remembering (kid's name, anniversary, allergy)
- The last time you talked and how it ended
You'll see a small memory tag above the draft showing where the context came from. Tap to expand the memory. You can also forget specific things in one tap.
Example draft surface
Sure — your appointment is still Thursday at 2pm. I'll bring the proofs you asked for last week.
02
Business knowledge graph
Business plan. Upload your business as documents and your replies start citing your own data.
Drop in:
- Product catalog (CSV, JSON, or PDF)
- Menu or service list (PDF, image, structured)
- FAQs and policies (returns, hours, delivery)
- Pricing sheets and discount rules
- Booking rules (availability, slot length, buffer)
WhatsAssist builds a structured knowledge graph from your uploads. When a customer asks "do you have it in blue under ₹2,000?", the draft already answers — with the actual SKU and price from your catalog. Update anytime. New uploads merge with existing. Conflicts get flagged for you to resolve.
03
Suggest, don't send
The default mode is conservative on purpose.
When a message arrives, WhatsAssist writes a draft and places it above your keyboard in the messaging app. You see it before anyone else does. You can:
- Tap to send as-is
- Edit then send
- Swipe to dismiss
- Long-press to give feedback ("not my style", "wrong info")
For high-confidence, low-stakes replies, you can opt into "Quick send" — one tap fires the draft without opening the app. Off by default. Always reversible.
04
Cross-channel context
Most people don't message you in one place.
Sarah texted, then DM'd you on Instagram, then sent a WhatsApp follow-up. To you it's one conversation. To every other app, it's three. WhatsAssist stitches it together. When Sarah messages you anywhere, the draft pulls from everything she has ever said to you on any channel.
The memory tag tells you where the context came from, so you can verify before you tap send.
05
Voice trained on you
Generic AI replies sound like ChatGPT trying to be helpful. WhatsAssist sounds like you.
WhatsAssist learns from your actual sent messages over the first week. After ~50 sent messages, drafts start adopting your sentence length, your common phrases, your punctuation habits, your level of warmth. Different tones for different contact groups — a draft to your accountant won't sound like a draft to your friend.
You can correct the tone anytime ("too formal", "more casual", "shorter"). It updates within a few messages.
06
Privacy by design
Privacy is product design, not policy text.
- All ingestion happens on your device via Android notifications. We never read your screen, your keyboard, or your storage.
- Memory is encrypted at rest. Per-user encryption keys.
- LLM calls send only the minimum context needed. No identifiers, no analytics joined to your messages.
- Zero training. Your messages never enter any AI model's training set.
- One-tap full delete. Everything is gone, on-device and on our servers, within 24 hours.
07
What we deliberately don't do.
- We don't send messages without your tap.
- We don't do bulk outbound, blasts, or marketing campaigns.
- We don't auto-add contacts or scrape contact lists.
- We don't sell, share, or rent your data to anyone.
- We don't train models on your conversations.