Rocal – 100% offline: a private “second brain” on your device
100% local · zero data leaks

No internet needed. Zero data leaks.
A private AI document analysis tool that fits in your pocket

No communication with outside servers at all. Rocal is an Android app in which an AI model (such as Gemma 4) runs on its own inside your smartphone, handling everything from fast search of confidential documents (RAG) to advanced AI conversation, completely offline and securely.

See the Rocal overview
Rocal screen 1
Rocal screen 2
01 / Concept & Philosophy

Private intelligence, behind an absolute line of defense

Completely removes the risk of leaking what you type into cloud AI

Contracts, financial data, unpublished medical records, proprietary research notes. Sending these to a cloud AI service carries the risk of serious company policy violations and data leaks.
You no longer have to give up on AI, and fall behind in productivity, because of security rules.

Rocal Principle No. 1: “Every outbound path is cut off”

Imported documents, chat history, OCR data and all inference happen only within an isolated local area of your smartphone.
Not a single byte goes to ad tags, tracking or outside communication. The app is designed never to send information out.

02 / Core Technologies

Four finely honed local features

FEATURE 01

Fully local RAG (document search and summaries)

PDF, Markdown and text files are indexed and searched instantly inside your phone. Answers are pulled from documents hundreds of pages long in an instant.

FEATURE 02

Obsidian Vault import (Premium only)

Import the notes and folder structure you built in Obsidian as they are. Use them as local knowledge, fully private, without sending anything out.

FEATURE 03

Shared Model Architecture

Share one model file (.gguf) with the other apps in the series (Tracal/Lingcal). No more storing the same multi-gigabyte file twice.

FEATURE 04

Tuned for your device's RAM

Depending on your device's memory (6GB/8GB/12GB/16GB), choose the right model, from lightweight models built for speed to high-precision reasoning models.

03 / Use Cases & Plans

Quiet protection for your thinking, and plan comparison

Legal & Compliance

Legal professionals

Analyze NDAs and confidential litigation documents locally, all at once. Key risk points are extracted automatically without sending anything to the cloud.

Corporate Strategy

Executives and financial strategy

Safely summarize M&A disclosure documents and pre-earnings information subject to insider rules. Support fast decisions without leaving logs outside.

Medical & Research

Medical and academic research

Turn patient case notes and unpublished papers into a local database and search them fast. Analyze safely in line with personal data protection rules.

Feature / Plan Free (Lite) Paid (Premium)
Outside communication / ad tracking Zero (0 bytes) Zero (0 bytes)
Local AI models stored 1 model (swap freely) All 24 models stored, switch instantly
Document (RAG) limits 1 folder, up to 5 files (1MB per file) Multiple folders, more files (with limits)
Obsidian Vault import Not available Import all Vault folders (stays on your device, nothing sent)
  • *1 Except when downloading an AI model for the first time, neither the free (Lite) nor the paid (Premium) version sends a single byte for ads, tracking tags, analytics or any other outside communication. The app has no feature that outputs or sends data outside.
  • *2 The free version (Lite) lets you store and use one model recommended for 6GB/8GB of RAM. The paid Premium plan lets you store all 24 models, including large reasoning models recommended for 12GB/16GB, and switch between them instantly.
  • *3 An internet connection is needed only to download AI model files for the first time and to update them.

Your intelligence, in your hands alone.

Back to portal