LOCAL-FIRST AI WORKSTATION · WINDOWS

Your Personal AI,
Running on Your Machine

RobiClaw is a complete AI workstation for Windows. Download one installer, run Gemma 4 models locally via llamafile, and get 11 automation Runners for research, coding, writing, and more. No cloud. No API keys. No subscriptions.

Download RobiClaw Windows 11 · Free download How it works →

100%

Local Processing

API Costs

AI Runners

RobiClaw — Research Runner

Research the best local AI models for my project and save a report

● Research Runner active…

workspace/research/ai-models-2026.md

# Local AI Model Comparison 2026

## gemma-4-E4B-it-Q4_K_M (recommended)
- Size: 3.3 GB · Context: 128 k
- Runs on 8 GB+ RAM, CPU or GPU
- Best for: code, writing, analysis

## gemma-4-E2B-it-Q4_K_M (lightweight)
- Size: 1.8 GB · Context: 128 k
- Runs on 4 GB+ RAM
- Best for: quick tasks, low-RAM machines

Report saved to workspace/research/

WHAT YOU GET

Everything in One Installer

RobiClaw bundles llamafile (Mozilla's portable LLM runtime), a Node.js API, and a React UI into a single NSIS Windows installer. No Docker, no WSL, no manual setup. Double-click and you're running local AI in minutes.

🤖

llamafile Runtime

Powered by Mozilla's llamafile — a single executable that runs on any CPU or GPU. OpenAI-compatible REST API. No separate install, no version conflicts.

🦞

11 AI Runners

Research, Code, Content Creator, Document, Workspace, Memory, and more. Each Runner has focused prompts, tools, and workflows built in.

🔒

100% Private

Nothing leaves your machine. No telemetry, no cloud calls, no accounts. Your conversations, files, and models stay local forever.

🧠

Semantic Memory

Local vector search via sqlite-vec + nomic-embed-text. RobiClaw remembers across sessions — stored in a local SQLite database, never the cloud.

📁

Workspace & Files

Research reports, code, documents, and notes are saved to ~/Documents/RobiClaw/workspace — readable, editable, yours.

🔍

Web Search

Brave Search API (free tier) or DuckDuckGo instant answers — used only by the Research Runner when local context isn't enough.

⚙️llamafile v0.10.0 (Mozilla)

🦀Tauri 2 (Rust shell)

💎Gemma 4 (Google)

⚛️React + TypeScript

🗃️sqlite-vec (vector search)

🟢Node.js 20 API

SETUP IN MINUTES

How It Works

The installer handles everything. No Docker, no WSL, no PATH wrangling.

Download the installer

Download RobiClaw_1.0.0.exe from GitHub Releases and run it. NSIS handles the rest — no admin rights required for per-user install.

Run onboarding

RobiClaw detects your RAM and GPU, recommends the right Gemma 4 model, and downloads it from HuggingFace in the background.

AI starts automatically

llamafile (chat), llamafile (embeddings), and the Node API all launch as sidecars inside the Tauri shell — no terminal windows.

Pick a Runner and go

Select Research, Code, or Content Creator from the sidebar to activate focused AI tools, workflows, and starter prompts.

WHAT YOU NEED

System Requirements

llamafile runs on CPU — no GPU required. A GPU makes it faster.

Minimum (CPU only)

Windows 11 64-bit
8 GB RAM
5 GB free disk (models are separate)
Internet for initial model download

Recommended (GPU)

Windows 11 64-bit
16–32 GB RAM
NVIDIA RTX (8 GB+ VRAM) or AMD RX 7000
SSD with 10 GB free

No Docker. No WSL. No Python. No Node.js required on the host machine.

PRICING

Free to Use, One-Time Upgrade

RobiClaw is free and open-source. The Pro bundle adds priority support, a custom Runner starter pack, and early access to new features.

Free / Open Source

Forever free · MIT / Apache

✓ Full RobiClaw desktop app
✓ All 6 built-in Runners
✓ Unlimited local AI usage
✓ Semantic memory (sqlite-vec)
✓ Windows, macOS, Linux

Download Free

Your Personal AI, Running on Your Machine