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

# What is Fieldwork

> The AI platform for growing businesses — an agent that works with your company's real data and tools

Fieldwork is an AI platform built for growing businesses, with a particular focus on ecommerce. You connect the systems your company already runs on — your store, your accounting, your files and databases — and work with an AI agent that can actually read from and act on them, with your approval.

## What the agent can do

<CardGroup cols={2}>
  <Card title="Work with your data" icon="database" href="/connectors">
    Connect Shopify, Xero, Google Workspace, databases, and more. The agent reads live data and takes approval-gated actions.
  </Card>

  <Card title="Analyze and build" icon="terminal" href="/files-and-data">
    Every conversation gets an isolated workspace where the agent can run code, query your tables with SQL, and produce downloadable files.
  </Card>

  <Card title="Remember what matters" icon="book" href="/knowledge-base">
    A versioned knowledge base the agent reads and — with your approval — maintains, so answers build on your company's context.
  </Card>

  <Card title="Run unattended" icon="clock" href="/automations">
    Schedule agents and actions: morning digests, data syncs, scripts — every standing job reviewed by you before it runs.
  </Card>
</CardGroup>

## How it's organized

Everything in Fieldwork belongs to an **organization**. Conversations, connectors, the knowledge base, files, automations, and skills are all scoped to your organization and shared according to access controls you manage. You can belong to several organizations and switch between them.

## The approval model

The agent asks before it acts. Actions with side effects — running commands in the workspace, editing knowledge base pages, calling a connector action that changes something in an external system — surface as **approval cards** in the conversation showing exactly what will run. Reads are generally exempt; mutations wait for you. You can enable auto-approve per conversation when you trust the loop.

## Where to start

<Steps>
  <Step title="Create your account and organization">
    Follow the [quickstart](/quickstart) — a few minutes end to end.
  </Step>

  <Step title="Connect a data source">
    Add your first [connector](/connectors) so the agent has something real to work with.
  </Step>

  <Step title="Have a conversation">
    Ask a question you'd normally dig through a dashboard to answer. See [working with the agent](/conversations).
  </Step>
</Steps>