CrackI
Prepare for your next opportunity.
An interview-preparation product that helps candidates work toward clearer, more confident answers.
Overview
A universal, ambient AI copilot that lives on top of your operating system. Press ⌘⇧Space from any app: CrackI reads your screen in real time via OCR, understands what app you're in, and answers anything in under a second. Multi-agent reasoning, persistent vector memory, live meeting copilot, and stealth mode (invisible to screen capture), all wrapped in a pill-shaped overlay that ships across macOS, Windows, Linux, iOS, Android, and the web.
The problem
Knowledge workers juggle 10-20 apps and 3-6 AI tools daily. Each is a silo: no shared context, no memory across surfaces, no awareness of what's actually on the screen. You re-explain, re-paste, and re-prompt every time, paying a context-switch tax all day long.
How it works
One ambient AI overlay summoned with one keystroke, working across every app on every device. Sees the literal pixels via PaddleOCR + app-context detection. A supervisor model decomposes complex queries into specialist agents (coding, research, summarizer, screen reasoner) and synthesizes one polished streaming answer. Every screen + conversation is embedded into a per-user Qdrant vector store, so CrackI remembers the Postgres schema you debugged last Tuesday and the customer you talked to yesterday.
What it does
- Pill-shaped overlay summoned by ⌘⇧Space, invisible to screen capture (stealth mode default-on)
- Sub-770 ms time-to-first-token from any app on macOS, Windows, Linux
- Multi-agent reasoning: planner → coding / research / summarizer / screen-reasoner → synthesizer
- Persistent per-user vector memory shared across desktop, web, mobile: one account, one memory
- Live meeting copilot: captures both mic and system audio, drafts answers in real time (interview / sales / support / coding personas)
- Smart model routing: Groq Llama for fast factuals, Claude for reasoning, GPT-4o for vision, local Ollama for sensitive queries
- Voice mode with push-to-talk and 80 ms interrupt handling
- Team workspaces: SSO (Google Workspace, Okta), audit logs, configurable data residency (US / EU / Asia)
- Free tier (30 queries/day) → Pro $29/mo → Team $79/seat/mo → Enterprise (self-hosted, air-gapped option)
At a glance
- Time to first token
- 770ms
- Surfaces
- Desktop · Mobile · Web
- Pricing
- $0 → custom
Notes from the build
CrackI is currently in active development. Desktop v1 (overlay + screen capture + multi-agent + live meeting copilot), the Web app, and the iOS / Android beta are shipped. SSO, SOC 2 Type II, and explicit opt-in Agent mode are in progress this quarter. Full 26-page product specification published at cracki.ai. Founding pricing: first 1,000 Pro subscribers get $19/mo for life.
Built with
Tauri 2 (Rust + WebView) · React Native + Expo · Next.js 14 · FastAPI · Postgres · Qdrant · S3 · Claude 3.5 Sonnet · GPT-4o · Groq Llama 3.1 · Ollama (local) · Whisper (Groq) · ElevenLabs TTS · PaddleOCR · LangGraph