Every model you need, one beautiful interface
Lattice is a fast, modern interface to your personal fleet of compute nodes — chat, vision, and code, routed across any OpenAI-compatible backend. No model is hardcoded; add a node, it shows up.
Access is invite-only. Ask the operator for an account.
Chat
Reasoning, writing, research
Vision
Images in, insight out
Code
Ship with a real editor
Built for your fleet
The front door to your private compute.
Connect local GPUs, a central gateway, or cloud providers. Lattice keeps model switching, files, memory, and conversations in one place.
Auto-route
Use the right lane by intent.
Private sessions
Your data stays yours.
One endpoint. Your models.
Use the same client you already know. Route requests through Lattice and keep provider changes out of your app code.
curl https://your-lattice-host/v1/chat/completions \\ -H "Authorization: Bearer $LATTICE_API_KEY" \\ -H "Content-Type: application/json" \\ -d '{"model":"lattice-auto","messages":[...]}'lattice-auto stays on the private fleet with provider caching disabled. lattice-auto-cached is an explicit opt-in that may send prompts to operator-allowlisted ZDR providers and request their short-lived prompt cache.
Chat
Stream from any model on the fleet — reasoning split out, markdown, code, and copy built in.
Vision
Drop in images and reason over them with the vision-capable nodes, no extra setup.
Code
A real editor beside a coder model — write, refactor, and review with the right node.
Backend-agnostic
Routes across any OpenAI-compatible node — local fleet, a central gateway, or cloud.
Smart routing
Auto-picks a model by intent: vision for images, a coder for code, chat for the rest.
Private by default
Your own accounts behind an encrypted session. No third-party auth, no telemetry.