Privacy
Keep sensitive work close
Prompts, files, and local inference stay on the machine you already trust. Privacy is the starting point, not a premium feature.
Backend.AI GO
Private AI,Real Agents,Ready to Mesh
Run models on your own machine, keep sensitive work close, and let agents handle the busywork.
Start on one laptop, then connect multiple GOs into a mesh across your office when you need more reach.
macOS
Windows
Linux
v1.6.1
Free to start · No sign-in required · Your data stays in your hands
No server setup · No Linux admin
Why GO Clicks
A desktop AI app for people who want control without the fuss.
Agentic Platform
Orchestrate Autonomous Agency
Transform your hardware into a high-performance agentic hub. Leverage MCP, secure tool-calling, and local sandboxes to build and deploy agents that solve complex, multi-step problems with true autonomy.
GO Mesh
One office, one shared AI workspace
Link multiple GOs on the same network so one strong machine can serve the rest. No server rack, no ops detour, no Linux setup.
Built for a Desk, Not a Server Room
Strong enough for daily work. Light enough to live on your desk.
Backend.AI GO is for developers and power users who want local privacy, useful automation, and an easy path from one machine to many.
Privacy First
Keep your prompts, files, and models nearby.
Run local models on macOS, Windows, or Linux and keep sensitive work on hardware you already know. When cloud models make sense, add them without giving up a local-first flow.
Hardware-aware acceleration across Metal, CUDA, ROCm, MLX, and CPU paths helps each machine do its best work.
Cloud APIs and remote inference servers can join the same workspace instead of sending you into a separate toolchain.
Local and remote models share one interface, so the workflow stays easy to follow.
Default mode
Local firstAgentic Hub
More than a chatbot. An autonomous platform.
GO is the hub where agents live and act. Transform your hardware into a high-performance platform that controls local tools and data directly. It is built for true agency, not just conversation.
Secure local sandboxes let agents run code and edit files without compromising your machine's integrity.
MCP (Model Context Protocol) provides a standardized way for agents to securely access your local context and data.
The same agentic infrastructure works across local models, cloud APIs, and remote backends.
Local tools
20+One place for models, agents, and the work around them.
Browse models, run agents, watch system activity, and keep the whole workflow in view from one desktop app.
A chat view that can actually get things done
Move between local and remote models in a conversation space that's already wired into tools, sessions, and model switching.
Beyond a Single Machine
Turn your office into a quiet little AI cluster.
Start on one desktop, then spread the model mix across the room. GO Mesh lets nearby machines share what each one hosts, so the office feels like one broader AI workspace.
Step 1
Connect your GOs
Backend.AI GO instances on the same network can find one another and form a mesh on their own.
Automatic discoveryStep 2
Share models acrossyour GOs
Each GO can host different models and make them available to the others, so you do not need to load every model onto every machine.
Different models, shared across GOsStep 3
Keep it pleasantly simple
No Linux admin, no server setup, no change in how people work. It just feels like the office got a little more capable.
No extra server setupPick your build and get started
Latest desktop release for macOS, Windows, and Linux. v1.6.1
Prefer Homebrew?
Install it in a single line
A tidy setup for macOS developers
Terminal
$brew tap lablup/tap
$brew install --cask backend-ai-go
Requires Homebrew. The cask tracks each stable release.
GO Enterprise keeps the local models, agent workflows, and GO Mesh setup your team already uses. It adds the rollout, control, and support organizations need.
Functionally, it is still Backend.AI GO. Enterprise changes how it is deployed, managed, and supported, not what your team has to relearn. Also available as part of the PALI suite.
Enterprise-ready additions
Air-gapped updates
Secure model and software delivery
Custom-branded builds
Prepared for internal distribution
Organization-wide routing
Shared policies across your deployment
Backend.AI Enterprise integration
Move into centralized AI operations