Trust by default
Most assistants treat your data like fuel. ZHAR treats it like a private conversation. Your conversations are encrypted with AES-256 on your device — we literally cannot read them.
Built in Berlin. Privacy first.
The AI that grows with you, without exposing you.
ZHAR is a personal AI that remembers your conversations for years, builds new skills from your requests, and stores everything encrypted on your device — never on our servers. Currently in private beta.
-
Privacy by architecture
No ad model. No data resale.
-
Persistent memory
Learns your context over years.
-
Evolution engine
Builds skills from conversation.
Why ZHAR
Built for people who want useful intelligence without surveillance.
Memory that matters
ZHAR remembers preferences, projects, and patterns so you do not repeat yourself every week. One thread, years of context, all stored locally.
Capabilities that evolve
Ask for what you need in plain language. ZHAR can turn repeated workflows into reusable capabilities — built from how you actually work, not a plugin catalogue.
How ZHAR compares
| Feature | ZHAR | ChatGPT | Claude | Siri |
|---|---|---|---|---|
| Where data is stored | Your device (AES-256 encrypted) | OpenAI cloud servers | Anthropic cloud servers | Mostly on-device; some cloud |
| Long-term memory | Months & years, on-device | Yes — cloud-stored, manually managed | Within Projects only, cloud-stored | Limited context from your apps |
| Builds new skills from conversation | Yes — from chat, automatically | No (Custom GPTs need manual setup) | No (Agent Skills need manual SKILL.md files) | No (Shortcuts only, manual) |
| Data used for training | Never | Yes, unless you opt out | Yes, unless you opt out | Yes, opt-out available |
| Open source (core) | Yes (planned) | No | No | No |
ZHAR in action
Real screenshots from the private beta. No staging, no stock photos.
How it works
One conversation thread. Clear context over time. No engagement optimization loops.
1. You talk naturally
No dashboards. No setup maze. Just describe what you are trying to do and where you are stuck.
2. ZHAR builds context
It remembers preferences, projects, and decisions so your next conversation starts from understanding, not from zero.
3. ZHAR evolves with you
Repeated tasks can become reusable capabilities. The assistant gets better because your work gets better.
Your work is confidential. Your AI should be too.
Client notes, privileged details, personal context — you need an assistant that can reference six months of conversations without sending a single word to someone else's server.
- "Summarise everything I discussed with Reyes since January."
- ZHAR pulls context from months of conversations — entirely on your device.
- Draft sent. Nothing logged, nothing trained on, nothing shared.
Lawyers, doctors, advisors, founders — if your work involves other people's trust, this is built for you.
Why this feels different
| Typical assistants | ZHAR | |
|---|---|---|
| Memory | Forgets context between chats. | Remembers threads and projects over time. |
| Privacy | Cloud logs, opaque retention. | Encrypted on your device. No ad model. |
| Workflow | One-off prompts and answers. | Builds reusable evolves from what you already do. |
Convinced? Reserve your spot in the private beta.
Get early access
The Evolution Engine
ZHAR's most unique feature: it turns your repeated workflows into reusable skills.
"Every week I prepare the same kind of client briefing — pull context from old conversations, check for conflicts, draft the summary. It takes an hour each time."
1
ZHAR detects a pattern in how you work
Not a single request — a recurring workflow that's costing you real time every week.
2
It proposes the "Client briefing" evolve
A reusable skill that pulls from your full conversation history, applies your formatting preferences, and knows which details matter to you.
3
"Prepare the briefing for tomorrow's call with Reyes."
One sentence. ZHAR drafts a structured brief — with context from six months of conversations, on your device, in seconds.
This is not a plugin marketplace. Evolves are born from how you actually use ZHAR — the requests you repeat, the patterns you don't even notice. No templates, no configuration screens, no code.
And because every evolve runs locally, the skill you build today stays yours — encrypted, portable, and invisible to everyone except you.
What we are building
AI should reduce cognitive load, not increase dependency. It should protect boundaries and respect attention.
ZHAR is built for that standard: a personal assistant that stays useful, remembers what matters, and keeps data under user control.
"ZHAR lights your path. It does not own it."
Product principle
Roadmap
A practical path from private beta to an open ecosystem.
Now — Phase 1
Private core
Personal assistant foundation with long-term memory and secure local-first architecture. Focus: trust and daily usefulness.
Beta access: Q2 2026
Next — Phase 2
Evolves engine
Turn repeated workflows into capabilities: email ops, scheduling, and structured personal automation from natural requests.
Target: Q3–Q4 2026
Later — Phase 3
Open platform
Open-source modules, community contributions, and transparent implementation patterns for privacy-first AI systems.
2027
Evolves
Product updates, new evolves, and curated industry news as they are released.
Built by a real person
ZHAR is built by Oleg Butov, a founder and engineer in Berlin. After a decade of shipping tools for teams buried in notifications and fragmented workflows, he built ZHAR to answer a simple question: what if your assistant actually remembered you — without that becoming a privacy problem?
Reserve your spot
Beta access opens Q2 2026. Join the waitlist and we'll reach out when it's your turn — one clear update, nothing else.
Join privacy-conscious professionals already on the list.