Sentinel Docs

Sentinel is a desktop app for working with AI inside a real codebase.

It helps to read the product as a work surface, not as a chat app with a few extra panels around it.

The app assumes there is a real project under the work. It assumes repo state matters. It assumes you might need a terminal, a diff, a browser tab, a checkpoint, or a follow-up message while a run is still going. A lot of the product only makes sense from that angle.

These docs follow the same shape. The early pages explain what the product is trying to do and why the app is arranged the way it is. After that, the docs move into setup, day-to-day workflows, runtime detail, and reference material.

If you're new here

Start with What Sentinel is. That page gives the shortest honest read on what the app is for.

Then read How Sentinel is shaped and Workspaces and threads. Those two explain the product model the rest of the app leans on.

If you already know you want to try it, jump to Quickstart. That gets you from install to the first real thread without much detour.

The sections

Overview

Overview is where the product model lives. These pages explain what Sentinel is trying to hold together and why the UI looks the way it does.

Getting started

Getting started is about the first hour with the app. It covers local setup, first launch, and the first useful workspace and thread.

Workflows

Workflows covers the part that matters once you are past setup. This is where thread behavior, plan mode, repo handoff, approvals, terminal use, browser use, and automations start to fit together.

Configuration

Configuration explains what can be connected, what can be turned on or off, and what the app stores. Provider setup, model setup, integrations, MCP servers, memory, and media features all live there.

Engines and runtime

Engines and runtime is the lower-level section. It is for people who want to understand how Sentinel carries state around a run and how Codex, Claude, and Copilot sit inside the app shell.

Product decisions

Product decisions is the reasoning section. These pages are less about usage and more about why the product ended up with this shape.

Reference

Reference is the compact lookup section. It keeps the tables, environment details, and release notes straight without turning everything into a tutorial.

A good reading order

1. What Sentinel is
2. How Sentinel is shaped
3. Quickstart
4. Workspaces and threads
5. Chat and plan mode
6. Repo workflow
7. Providers and models
8. Sentinel engine

Other docs in the repo

There are also a few practical reference pages that are less about product behavior and more about shipping and maintaining the app: Release process, macOS signing, and Desktop platform smoke checklist.