Skip to content
Settings

Follow your day.

Local clock approximation

Location stays in this page’s memory and clears on navigation. This imagined west-facing landscape is not local weather. The sun or moon may be outside the frame.

Scene lighting follows the selected mode. Home plots the astronomical sky; product scenes are cinematic illustrations with animated sky and optional water, not a view of your actual surroundings.

Real field recordings blend with the scene: birds by day, quieter evenings, soft insects at night. Water sound is separate from water motion. Audio pauses when you leave this tab. Recording credits.

HT Studio

Local intelligence.
Naturally yours.

Your models. Your machine. One open interface.

Explore the runtimeView source
DaemonCLISDKOpenAI-compatible API

One ecosystem.
A clear path.

From local models to real applications.

01 · Choose a model

Find a fit.

Compare model options and runtime requirements in Marketplace.

Explore the catalog 02 · Make it local

Use your hardware.

Manage the model and check your workstation’s readiness in HT Studio.

Explore the runner 03 · Connect your work

Build something useful.

Connect an app or agent through the local API. Talk through an integration.

Talk about an integration

Less infrastructure.
More room to build.

A small control plane for the heavy work of local AI.

HT Studio manages a native llama.cpp engine, model inventory, and workstation readiness. Your application talks to one familiar interface instead of managing every model-specific detail.

It’s the runtime—not a marketplace or a bundled collection of model weights. The models and dependencies live on your machine.

See what’s
actually running.

Real terminal capture from the standalone build. Play it when you’re ready.

HT Studio · build evidenceLocal runtime

Recorded build evidence · June 2026 · not a live session on your computer

Keep models close.

A warm set supports switching between models without starting a fresh runtime for every request.

Work with your hardware.

Hardware-aware launch settings and CPU offload support models that don’t fit entirely in GPU memory.

Make the connection.

Use the local OpenAI-compatible endpoint or typed SDK. The default runtime path stays on the workstation.

From model to application

A familiar interface.
A local foundation.

Connect an app, agent, or tool without giving it responsibility for the entire runtime.

  1. Start your runtime.
    htstudio serve
  2. Check local readiness.
    htstudio status
  3. Connect through the local API.
    http://127.0.0.1:3001/v1
  4. Use the SDK for typed daemon calls.
    Read the source and integration docs

Small enough to inspect.
Open enough to trust.

Keep the beautiful presentation separate from the evidence. Inspect the captured output and the source behind it.

Inspect the build footprint Source, not model weights
ScopeRecorded sizeWhat it includes
Core code, docs & config1.02 MBExcludes media, dependencies, dist and output
Runtime packages source474.84 KBSDK, daemon and CLI
Studio app source101.01 KBFocused management interface
Tracked with proof media4.84 MBRepository payload with screenshots and video
Open the evidence gallery 4 original captures
HT Studio terminal output: Footprint proof
Footprint proof
HT Studio terminal output: Smoke checks
Smoke checks
HT Studio terminal output: CLI surface
CLI surface
HT Studio terminal output: Live daemon
Live daemon
Open raw proof video

Your next idea.
Powered by your machine.

Explore the open source, or talk with Hass about bringing local AI into your workflow.

Explore the sourceTalk about an integration