A control surface for your own AI workstation

Your compute path is your decision.

Keep compatible inference, files, and tools on your machine when you configure a local route. Use hybrid or hosted paths only when the work calls for them.

Buy Telethryve Desktop — $19.99 Download Local Stack 0.1.16

Local compute

Local compute starts with your machine.

A compatible local model runtime can run inference beside the files, tools, and task state already on your workstation. Telethryve gives that setup a visible control surface instead of making the route an assumption.

Keep work closeCompatible inference, project files, local tools, and task state can remain on the workstation.

Choose a local profileSelect a local profile in Telethryve and work from the same project surface. Hardware and runtime compatibility determine what can run locally.

Use hosted capability on purposeWhen a task needs Claude, Codex, or another configured API, choose that route deliberately. Only that selected task crosses a provider boundary.

Local Stack 0.1.16 for Apple silicon

The signed and notarized Local Stack package installs the Telethryve Codex engine and local Gemma profile. Ollama and the Gemma model remain local dependencies; the installer verifies readiness before marking the stack ready.

The value

The value is control: fewer unnecessary handoffs, clear privacy boundaries, and a deliberate trade-off between local capability, speed, and cost.

Local and hosted paths can work together. Provider accounts, settings, and terms still govern any provider route you select.