horsie server

Run LLM agents as durable chat sessions in your browser, with tools executing on a machine you chose.

Start a session

Install the server, add a model, give it somewhere to run tools, and chat. Quickstart

Choose where tools run

On your own machine over a dialed-out connection, or on containers the server creates for each session. Runtimes

Run work unattended

Save an agent and a prompt as a routine, or chain agents into a workflow graph. Routines · Workflows

Understand the design

Why sessions survive a restart, and why runtimes dial out instead of being dialed into. How it works