“Explain this function.”
A lightweight local model can handle a simple request.
LOCAL MODELLOCAL FIRST. YOUR CONTROL.
Your AI tools.
One gateway.
Install LiteGate on your computer and connect your AI tools. It directs their requests to your local or cloud models, following your preferences.
In development · Planned for macOS, Windows and Linux
HOW IT WORKS
Install. Connect. Choose.
The experience we are building. Public installers are not available yet.
Install it on your computer.
LiteGate runs locally. It becomes the connection point between your AI tools and the models you choose.
Connect your tools and models.
Point compatible assistants at LiteGate. Add your cloud providers and the local model engines you already use.
Choose how requests are handled.
Select a model yourself, or let routing consider task complexity, availability, privacy and your configured limits.
AN EVERYDAY EXAMPLE
Keep using your editor. Let LiteGate choose the route.
Imagine building a website in Visual Studio Code with an AI assistant such as Continue configured to send its requests through LiteGate.
“Help me solve this complex bug.”
LiteGate can select a more capable cloud model, when your rules allow it.
CLOUD MODEL“Can I keep working?”
New requests can use a compatible local model you have running, if the route allows it.
LOCAL ALTERNATIVEOne assistant. One connection. Your models, under your rules.
Illustrative scenario; the Continue integration still needs version-specific validation. LiteGate routes only requests sent to it and does not install local models. Cloud subscriptions require a supported provider integration, still in development.
Read about connections and requirements →GATEWAY + CONTROL CENTER
Your AI setup, in one place.
Desktop, chat, routing and LAN implementations are integrated. Public release validation is still pending; expanded telemetry and subscription inventory remain planned.
Configure once. Stay in control.
Manage providers, credentials, routes and diagnostics from a native desktop app. Keep the gateway running from the menu bar or tray, even when the chat window is closed.
Choose how each request travels.
Manual keeps the choice with you. Smart ranks eligible models locally. Private prefers local models and permits explicitly approved remote providers. Local excludes every remote destination.
See what each request costs.
Inspect tokens, estimated cost and routing decisions. The product plan extends this view with time to first token, tokens per second, latency and a separate subscription inventory.
Your gateway at a glance
Check activity, token usage and the routing mode from one control panel.
Development version · Demo data · Interface shown in English
Select the image to enlarge ↗LITEGATE IN YOUR WORKFLOW
Your tools. A shared set of rules.
Product scenarios · Public release pending. Client compatibility depends on protocol, version and enabled capabilities.
Codex, on your terms
Connect Codex to one stable endpoint and manage destinations and policies through LiteGate.
Explore this use case →Claude Code, connected
Work in your repositories with centralized provider credentials and shared policies.
Explore this use case →A backup plan for your models
When a recoverable failure or provider limit occurs, try another allowed destination before delivering the response.
Explore this use case →Your tokens, your budget
Set budgets per tool, project or key, with visible usage, alerts and explicit policies when a limit is reached.
Explore this use case →Your own chat, the same rules
Talk to your models inside LiteGate, with local history and visible routing decisions.
Explore this use case →Your models, from another device
Share access with trusted devices through LAN, TLS and keys with scoped permissions and limits.
Explore this use case →Local and cloud, in the same place
Give your models stable names and inspect the destinations in each route.
Development version · Demo data · Interface shown in English
Select the image to enlarge ↗A chat of your own
Keep conversations and model selection together, with local history and visible routing controls.
Development version · Demo data · Interface shown in English
Select the image to enlarge ↗HOW THE PIECES FIT
One connection.
You choose the path.
Your tools talk to LiteGate. Your models stay with their local engines or cloud providers. The gateway is designed to connect them under your rules.
Explore the planned architecture →ONE CONNECTION. MORE POSSIBILITIES.PRODUCT VISION
CodexClaude CodeYour app
↓↓↓
LiteGateYour AI gateway
PrivacyRoutingVisibility
↓↓
Local modelsOn your machine
Cloud modelsWith your providers
Your models run in their own engines.
RELEASE READINESS
Built. Tested locally.
Preparing for release.
Gateway, desktop, chat, Smart routing and LAN implementations are integrated. Signing, final native platform checks and clean-machine release validation remain open. No public installer is available yet.
Check the project status →START HERE