YandexGPT · GigaChat · Claude · GPT — switch in one click

Real-time AI hints
for calls and interviews

The app listens to the conversation and within 1–2 seconds shows a ready answer. The window never appears in screen capture — the other person sees nothing.

Loading…See use cases
100 ₽ sign-up bonus No subscriptions Pay with a Russian card
14:32·Tue, May 19Wi-Fi 87%
Recording · Zoom · Avito · Senior Backend interview
HR · Avito
You
Speaking
End
Invisible on the shared screen1,4s
"Tell me about your experience with microservices and Kubernetes?"
AI · Claude Sonnet 4.6
3 years: a platform on K8s — 40 microservices. Helm charts for CI/CD, deploys cut from 25 to 4 minutes. Istio for the service mesh, Prometheus + Grafana.
~50 ₽
per hour — typical spend
6 models
AI to choose from in one window
16 languages
recognition & answers
100 ₽
bonus for new users
< 2 s
average AI response time
Who it’s for

One service — four scenarios.

Pick your case. Hints are tailored to the scenario, attached files, and answer style.

Most common case
Job candidates

Land the offer you’ve worked toward for years.

Senior and lead positions, IT, product management, finance. You get one shot at an interview — don’t lose it to stress or a forgotten formula.

  • Hints based on your resume and the job posting
  • System design, technical questions, behavioral
  • Plug in Claude or GPT-4 for complex topics
8 of 10
candidates got an offer from the first interview
JC
James Carter
Senior Backend, moving from e-commerce to fintech
Resume 2026.pdf
Job description — Stripe.pdf
Interlocutor
Tell me, how would you design a system handling 10M RPS?
AI · 1.4s
I’d start with three layers: an L4 load balancer (nginx/envoy), then an API gateway with rate limiting. Stateless nodes scale horizontally. For hot data — a Redis cluster sharded by user_id. Cold data — PostgreSQL with read replicas and Citus. Queues on Kafka, eventual consistency for most operations.
Claude Sonnet 4.6 · 3 ₽ / 1k tokens~80 ₽ for an hour-long interview
How it works

From “never tried it” to your first hint — 5 minutes.

01
1 minute

Sign up

By email. You get 100 ₽ as a bonus — enough for a typical interview or 8 hours of calls.

02
20 seconds

Install

Download a regular exe installer. No VPN, no admin rights.

03
Before the call

Launch

Attach files (resume, price list, notes). The window appears over Zoom — and isn’t visible during sharing.

04
In real time

Hints

The AI hears the question and shows an answer within 1–2 seconds. Read it aloud or use it as a guide.

AI models

6 models in one window. Price shown before the request.

Switch in one click. All models work without a VPN. Prices are per 1,000 answer tokens.

ModelPriceLatencyRussianContextWhen to choose
GPT-5 nano
OpenAI
0,070 ₽/ 1k0,7 сgood400kCheapest and fastest — for calls
GPT-4o mini
OpenAI
0,11 ₽/ 1k0,9 сgood128kBudget option for calls
GPT-4o
OpenAI
1,8 ₽/ 1k1,3 сexcellent128kVersatile complex questions
Claude Haiku 4.5
Anthropic
3,2 ₽/ 1k0,9 сgood200kFast answers, economical use
Claude Sonnet 4.6
Anthropic
8,5 ₽/ 1k1,6 сexcellent200kTechnical interviews, system design
Claude Opus 4.8
Anthropic
16,7 ₽/ 1k2,2 сexcellent200kTop quality, complex tasks
The output price is shown BEFORE you send the request Yandex and Sber servers are in Moscow All models work without a VPN — via Russian infrastructure Recognizes speech and answers in 16 languages — with automatic language detection
Security and privacy

What the other person will never see.

No magic. A native Windows API, servers in Moscow, AES encryption of the local database.

Invisible during sharing

The window uses a native Windows mechanism (WDA_EXCLUDEFROMCAPTURE) that hides it from recording and streaming software at the OS level.

  • Yandex Telemost · Kontur.Talk · VK Calls
  • Microsoft Teams · Zoom · Google Meet
  • OBS · Discord · Win+Shift+S · screenshots

Audio is not stored

Sound goes to speech recognition only while the app is running. After you close it, no recordings or cache remain.

  • Recognition — Yandex SpeechKit
  • Not a single byte of audio is kept after the session
  • The stream is encrypted with TLS 1.3

History — on your PC only

Chats and attached files are stored locally in an encrypted database. The server only sees tokens and balance.

  • AES-256 for the local database
  • Delete the app — no data remains
  • Prompter Assist servers don’t store chats

Payments via YooKassa

Payment data is handled by YooKassa. We never see or store it. A receipt is emailed to you automatically.

  • МИР, Visa, Mastercard cards (RU)
  • PCI DSS Level 1 — on YooKassa’s side
  • Any transaction — a refund button in your account
Legally clean.
We operate under 152-ФЗ and НК РФ, with payment via YooKassa and an automatic receipt. Responsibility for use on a specific call rests with the user.
TermsPrivacy
Pricing

Pay only for what you used.

No subscriptions. Minimum top-up — 10 ₽. The 100 ₽ sign-up bonus is enough for a typical interview.

Calculator

Estimate your typical day

4 h
1 h (one interview)10 h (a full workday)
Spend per day
~332₽
GPT-4o · pay as you go by tokens actually used
Price includes speech recognition (~39 ₽/hour)
100 ₽
Sign-up bonus
Enough for one full interview or 3 hours of calls on a lite model
10 ₽
Minimum top-up
Via YooKassa, with a Russian card. Credited in 1–5 minutes
0 ₽
Subscription fee
No subscriptions, auto-charges, or mandatory packages
6
AI models to choose from
YandexGPT, GigaChat, Claude, GPT — switch in one click
Frequently asked questions

We answer straight.

Yes. It uses a native Windows mechanism (WDA_EXCLUDEFROMCAPTURE) that hides the window from recording and streaming software at the OS level. Verified in Yandex Telemost, Kontur.Talk, VK Calls, Microsoft Teams, Zoom, Google Meet, OBS, with standard Win+Shift+S, and in Discord.
By default the window pins next to the webcam — so your eyes stay “on camera.” You can adjust the font size and density to read one line at a time. Most users get used to it after 2–3 sessions and don’t look away.
Legally — no. Using external tools is not prohibited by Russian law. Many users treat Prompter Assist as a “second screen with notes” — it helps you not freeze up, not invent experience. If the experience is real, hints only help you phrase it correctly.
Windows 10 (version 2004 and newer) and Windows 11. macOS and Linux are not supported yet.
Speech recognition works in 16 languages: Russian, English, German, Spanish, French, Italian, Portuguese, Dutch, Polish, Finnish, Swedish, Turkish, Kazakh, Uzbek, Hebrew, and Ukrainian. You can enable auto-detection (up to 5 languages at once) and set the AI to reply in the other person’s language, in your own language, or as instructed by your prompt.
No. All models — YandexGPT, GigaChat, OpenAI, and Anthropic Claude — work without a VPN. Requests to OpenAI and Claude go through Russian infrastructure, so they’re available directly.
With a Russian bank card (МИР, Visa, or Mastercard RU) via YooKassa, in rubles. There’s no subscription — you top up an internal balance that’s charged as you use it. A receipt is emailed automatically.
Audio goes to recognition (Yandex SpeechKit) only for the duration of the conversation and is not stored. Message texts are sent to the chosen AI model only at the moment of the request. Chat history is stored only on your computer, in an encrypted local database.
100 ₽ bonus on sign-up

Your next call —
with hints.

Installs in 1 minute. No subscriptions or auto-charges. Pay only for the tokens you actually use.

Loading…Compare models
Windows 10/11 · 84 MB installer · pay with a Russian card