---
title: "Introduction"
description: "What Trevize is, how teams build software with AI agents, and why every change lands as a reviewable pull request."
---

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

# Introduction

Trevize is a shared workspace where your team and AI agents build software
together. You describe what you want, agents do the work inside isolated cloud
sandboxes, and every change comes back as a **pull request on your own GitHub
repository** — so nothing ships without your review.

## Two ways to work

Trevize gives you two complementary modes, backed by the same agents and the
same repositories:

- **Live sessions** — A real-time, multi-player working session. You and your teammates chat with
an agent as it works, steering it turn by turn — like pairing, but the agent
writes the code. See [Sessions](/concepts).
- **Automations** — Autonomous loops that run on a schedule and keep improving your product
between sessions, opening pull requests as they make progress. See
[Automations](/concepts).

For one-off work you don't need to sit through, you can also start a **task** — a
single unit of work that runs autonomously from a short prompt and a blueprint
(a bug fix, a feature, or a quick change) and produces a pull request.

## How it works

1. You **connect a GitHub repository** to a workspace.
2. You start a session, task, or automation and describe the goal.
3. Trevize spins up an isolated **sandbox**, checks out your repo, and runs a
   **Claude Code** agent against it.
4. The agent implements the change on a branch, optionally serves a **live
   preview**, and opens a **pull request**.
5. You review, leave feedback, and merge — on your terms.

> **You stay in control**
>
> Agents work on branches and propose pull requests. Trevize never merges to your
> default branch on its own — humans review and merge, exactly like any other
> contributor.

## What it connects to

Trevize meets your team where it already works:

- **GitHub** — repositories, branches, and pull requests.
- **Linear** — turn issues into work and track it across tools.
- **Claude Code** — the coding agent that powers sessions, tasks, and automations.
- **MCP**, **Neon**, **Sentry**, and **Supabase** — extend agents with tools,
  databases, error context, and backend services.

See [Integrations](/integrations) for the full list.

## Who it's for

Trevize is built for product teams who want to move faster with AI agents
without giving up code review, previews, or ownership of their repositories.
Work stays in your GitHub, agents carry [knowledge](/concepts) forward across
runs, and [insights](/concepts) show you exactly how much they're contributing.

## Next steps

- **Core concepts** — A fast tour of the vocabulary you'll see throughout Trevize.
[Read the tour](/get-started/core-concepts).
- **Quickstart** — Connect a repository and start your first session.
[Get started](/get-started/quickstart).

Source: https://docs.trevize.dev/get-started/introduction/index.mdx