Vibe code with local models
Local models have gotten good—fast. Download and run open-source models right in the app. Fully private, fully free, works offline. Use them on their own or alongside cloud models for the best of both worlds.
“For years local models felt like a science project—cool demo, unusable for real work. Then almost overnight they just started working. They're not perfect for everything and I still reach for cloud models on heavier tasks, but having AI run right on my laptop with total privacy and zero cost? It's hard to explain the feeling. My data never leaves my machine. I never worry about an API key leaking. It just feels like the future arrived quietly.”
Why go local?
Private, free, and actually useful—the trade-offs aren't what they used to be.
Complete Privacy
Your code never leaves your machine. No data sent to external servers, no API keys exposed to third parties—ever.
Free & Fully Offline
No API keys, no usage limits, no surprise bills. Turn off your internet and keep working—your AI runs entirely on your hardware, no connection required.
An Ecosystem on Fire
QWEN 3.5, Minimax, GLM, Neo-Metron by NVIDIA—capable open-source models are shipping every week. Workshop stays current with the latest so you always have the best options for your hardware.
Set up in minutes, code forever
From download to building—no accounts, no API keys, no config files.
- 1
Download the desktop app
Install Workshop Desktop for macOS, Windows, or Linux. It includes everything you need.
- 2
Pick and download a model
Workshop recommends models for your hardware—a fast model for quick tasks and a high-quality model for deeper work. It grades whether your machine can handle fully agentic coding or is better suited for chat. One click to download, zero setup.
- 3
Start building
The same chat-to-build workflow as the cloud version—but everything runs locally. No API keys, no usage limits, no internet required.
- 4
Go hybrid
Run local and cloud models side by side. Use local for privacy-sensitive work and cloud when you need maximum horsepower. Power users can also connect Ollama, LM Studio, or any Anthropic-Messages-compatible endpoint.
“The game changer for me was realizing I don't have to choose. I run local models for anything sensitive—client work, internal strategy, anything with credentials—and cloud models when I need the extra horsepower. Having both side by side in Workshop means I'm always working. One thread is fully private on my machine, the next one taps Claude for something complex. It's the most productive setup I've had.”
Built for teams that need privacy
Private Codebases
Work on proprietary code with AI assistance. No snippets or context ever leave your machine.
Regulated Industries
Healthcare, finance, and government teams can use AI coding while maintaining full compliance.
Offline Development
Code anywhere—planes, trains, remote locations. Your AI assistant travels with you.
Air-gapped Environments
Deploy in fully isolated networks where cloud services are not an option.
Low-latency Workflows
Local models respond instantly with zero network overhead. Ideal for rapid iteration loops.
Self-hosted Teams
Organizations that self-host their tooling can add AI coding without changing their security posture.
Your code. Your machine. Your rules.
It's free to try, no credit card required.