> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orbeex.app/llms.txt
> Use this file to discover all available pages before exploring further.

# How Orbeex Works

> Four simple steps from your request to a finished result.

# How Orbeex Works

Orbeex works like a careful assistant who shows their work before touching anything.

## The four steps

### 1. You ask

Type what you want in everyday language, or use the normal menus and tools.

> "Find every building inside the flood zone and export the list."

### 2. Orbeex plans

For larger requests, Orbeex breaks the task into clear steps and shows you the plan first:

1. Load the buildings layer
2. Load the flood zone layer
3. Select buildings inside the zone
4. Export the result table

### 3. You approve

You review the plan. Change it, ask questions, or let it run. Risky actions — anything that overwrites or deletes data — always require your explicit confirmation, and you can see a preview first.

### 4. Orbeex executes

The work runs on your computer. Each step is checked and recorded.

## Everything is recordable and reversible

This is the part that makes Orbeex trustworthy:

* **Every action is saved as a step** in a running history.
* **Undo anything** — step back to any point, like undo in a text editor but for your whole analysis.
* **Repeat exactly** — run the same sequence again on new data and get the same result.
* **Prove it later** — show a client or auditor precisely what was done and in what order.

This history is stored locally, in your project file.

## Where the AI fits in

The AI is the part that understands your plain-language requests and proposes the plan. It never touches your data directly — it only suggests steps, and the engine checks and runs them. That separation is why AI suggestions cannot silently misbehave: the engine enforces the rules, and you hold the approval.

More detail: [Using AI with Orbeex](/working-with-ai) and [Safety and guardrails](/ai/safety).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.