Local and intranet models
Manage local inference processes or register an existing enterprise model endpoint. Assign the connection to eligible Harnesses.
INTELLISYS / CONFIGURATION
Models & skills
Shared model connections, local inference controls and skills for employee environments.
THE MODEL SERVICE LAYER
Three sources.
One place to organize them.
Manage local and public model connections alongside vendor-backed CLI environments. Each source keeps its own authentication and authorization method.
MODEL SERVICE PATHS
Full size ↗Local models, public APIs and vendor services
CONFIGURATION AND AUTHORIZATION
Central configuration.
Direct model requests today.
Administrators maintain connections, choose defaults and assign eligible employees. The Harness sends inference requests to the actual endpoint.
Public model APIs
Register endpoints, protocols, keys and model lists. DSH receives settings through its API; pi receives configuration files.
Vendor-backed CLI tools
Manage the installed version, container and terminal entry. Employees currently sign in to Codex CLI and Claude CLI with their vendor accounts.
CURRENTConnections · assignment · runtime management
ROADMAPOptional proxying · usage reporting · spending controls
LOCAL INFERENCE
From hardware to a model service.
CroGate manages independent inference processes on the host. The engine serves model requests on its own endpoint.
01
Detect
Inspect the host’s hardware and available resources.
02
Prepare
Install an engine and acquire or register a model.
03
Run
Create the service, set parameters and observe its state.
04
Assign
Register it as a shared connection for eligible environments.
| Engine | Architecture scope |
|---|---|
| llama.cpp | CPU and supported accelerator paths |
| vLLM | Supported Linux GPU deployments |
| SGLang | Supported Linux GPU deployments |
| oMLX | macOS engine path; Mac server expansion is planned |
SHARED WORKING METHODS
Package the method.
Deliver it where people work.
Skills contain instructions, scripts and reference material. They are managed separately from model services and combined by the runtime during work.
SHARED SKILLS
Full size ↗Make working methods reusable
Create
Write a skill or paste an existing method into the shared library.
Upload or import
Bring files and folders, or import a snapshot from a server directory.
Assign
Choose the employees who receive the skill and follow delivery status.
KEEP EXPLORING