Destaris
Browse the docs

Introduction

What Destaris is, and why it's shaped the way it is.

Destaris is a desktop app for building and running automation workflows — deterministic by default, AI opt-in per step, executed on your own machine. You compose a workflow on a visual canvas, it runs locally on a schedule, and a dashboard shows you what ran, what it cost, and what broke.

The shape of a workflow

Every workflow is a small graph of steps:

  • A trigger decides when it runs (a schedule, or you run it by hand).
  • Deterministic steps do the plain work — fetch data over HTTP, transform it, branch on a condition, cache what's already been seen. They do exactly the same thing every time and cost nothing to run.
  • AI steps are added where you actually need judgment — and nowhere else. They run as your own AI CLI, on your own plan.
  • Outputs send the result somewhere (post to an API, notify a channel).

The same workflow is just as readable as portable YAML — the canvas and the file are two views of one thing.

Three ways to automate

A workflow is the first of three process objects, and the app offers all three side by side. They're ordered by how much of the process the AI owns:

  • Workflow — for steps you know. A fixed set of steps you build once. It runs the same way every time, for next to nothing.
  • Loop — for steps that take a few passes. You build the steps, same as a workflow, and you also write down what "done" means. Each pass, a small brain checks the result against that and sends specific stages back to try again — until it passes.
  • Agentic Workflow — for figuring it out. Give it a goal instead of steps. A brain plans, runs, and adapts until the goal's met, building a workflow as it goes.

The split that trips people up isn't a single scale. A workflow and a Loop are authored the same way — you build the steps in both — and a Loop just adds a definition of "done" and reruns itself until that's met. An Agentic Workflow is the one where you don't author the steps.

An Agentic Workflow works itself out of a job: once it does the same thing the same way, it settles into a plain workflow, and you stop paying for the brain. So does a Loop, on a stricter bar.

Why it's built this way

Most "AI workflow" tools run the AI in their cloud and bill you per run. Destaris runs it on your machine with your logins, keeps deterministic steps free, and never meters your runs. Three ideas carry the whole product:

  • Deterministic-first. Intelligence is something you opt into per step, not the unpredictable default.
  • Local brain. Execution happens on your computer; nothing is custodied.
  • No execution tax. You pay a flat price (and the desktop app is free) — never per run, because your machine does the work.

Where to go next

  • Quickstart — install Destaris and build your first workflow.
  • Deterministic-first — the core idea, in depth.
  • The canvas — how to build workflows visually.
  • Loops — the same steps, plus a definition of "done" they have to clear before the job counts as finished.
  • Agentic Workflows — hand over a goal instead of steps.