Local inference · Central control

Your models. Your hardware.
One console.

Run the Layman Agent on any machine to serve an OpenAI-compatible API from local models — then manage the entire fleet, hot-swap engines, and broker cloud fallback from Strata Console.

Detecting your platform…

OpenAI-compatibledrop-in /v1/chat/completions
Hot-swap enginesOllama · llama.cpp · MLX
mTLS enrollmentper-agent X.509 identity
Cloud fallbackbrokered Anthropic / OpenAI
How it works

Two pieces, one system

Install the agent where your models live. Manage everything from the console.

1

Download the agent

Grab the Layman Agent for your OS and run it on the machine with your GPU or CPU inference stack.

2

Enroll into the console

The agent dials out to Strata Console and enrolls with a per-agent certificate. No inbound ports to open.

3

Manage the fleet

Set the engine, model, and guardrails from the console. Route traffic and fall back to the cloud on demand.

Layman Agent

A lightweight runtime for local models

A single Go binary that exposes an OpenAI-compatible inference API and a control API the console drives. Swap the model or the entire engine on demand — but only when no request is in flight.

  • OpenAI-compatible /v1/chat/completions, /v1/models, /health
  • Engine adapters: Ollama, llama.cpp (GGUF), MLX
  • Idle-safe swap coordinator with optional drain
  • Local status UI with live download & lifecycle events
  • Outbound-only connection to the control plane
Strata Console

The control plane for your fleet

A multi-tenant management plane for fleets of Layman Agents. Enroll agents with workload identity, push configuration, route requests, broker cloud fallback, and meter usage — all from one place.

  • Per-agent mTLS enrollment (SPIFFE-style X.509 identity)
  • Catalog & groups, scoped enrollment tokens, revocation
  • Central config: engine, model, params, system & guardrail prompts
  • Cloud fallback brokering to Anthropic / OpenAI
  • Usage metering across the whole fleet
Download

Get the Layman Agent

Pick your platform. Version .

Installing on a fleet? After download, run the agent with -control-mode remote and enroll it from Strata Console.

Ship local inference, managed centrally.

Download the agent, spin up the console, and bring your fleet online.