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

# How OnePatch works

> Telemetry in, an agent over your data, monitors and dashboards out.

OnePatch has three parts: the telemetry you send, the agent that reads it, and the artifacts the agent produces. This page explains the model so the rest of the documentation reads clearly.

## Telemetry in

Your application sends OpenTelemetry — traces, metrics, and logs — to your workspace's ingest endpoint. OnePatch stores every signal and makes it queryable. You do not configure storage, schemas, or retention; you point your exporter at the endpoint and the data lands.

This is the only thing you have to set up by hand, and you do it once. See [Getting started](/getting-started) for the steps.

## The agent over your data

Your workspace runs an AI **agent** with read access to all of your telemetry and to the repositories you connect. You interact with it in chat. It is the single interface to the product — there is no separate query builder, alert editor, or dashboard designer.

When you ask a question, the agent queries your telemetry directly and answers from the data. When you ask it to watch something or visualize something, it writes the monitor or dashboard for you.

<Note>
  The agent answers from your actual telemetry, not from a fixed set of pre-built reports. If your service emits it, the agent can query it.
</Note>

## Monitors and dashboards out

The agent produces two kinds of durable artifact:

* A **dashboard** is a live view of your telemetry. The agent writes it, and it refreshes as new data arrives.
* A **monitor** is a small program that checks a window of your telemetry on a schedule and reports whether it is `healthy`, `firing`, or `unknown`.

Both are stored as code in your workspace and versioned in git, so every change is a commit you can read and revert. A monitor goes live when the agent commits and pushes it — there is no separate publish step. [Monitors and dashboards](/monitors-and-dashboards) covers this in detail.

## What the agent does when something breaks

A monitor can be **armed**. When an armed monitor fires, OnePatch opens an incident and the agent triages it — it investigates the firing, and either tunes the monitor if it was a false alarm or opens a pull request if it found a real problem. Notifications land in your `#onepatch-alerts` Slack channel. [Incidents and Slack](/incidents-and-slack) explains the full path.

## Vocabulary

These terms appear throughout the product and the docs:

| Term              | Meaning                                                     |
| ----------------- | ----------------------------------------------------------- |
| Workspace         | Your organization's private OnePatch instance.              |
| Server            | The machine your workspace runs on, including the agent.    |
| Agent             | The AI runtime inside your server that you talk to in chat. |
| Online            | The state of a workspace that is provisioned and ready.     |
| Telemetry flowing | Your workspace has received its first signal from your app. |