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

# Run your ad accounts with AI agents

Adjar is the control layer for AI agents across your ad accounts. Your account becomes **plain-text config** the agent can read and change for you. It drafts the edit, you review the before-and-after, and Adjar applies it. Nothing goes live until you say so.

Adjar supports **Google Ads** and **OpenAI Ads** (the ads that run inside ChatGPT) today, through one workflow and one CLI.

## Why it exists

Ad consoles were built for human clicks. That's why your AI agent can't help with paid media the way it already helps everywhere else — it has no way to work a console.

Adjar gives the agent something it can work: plain text. Your campaigns, budgets, and creatives become files the agent reads and edits. Your performance report sits right next to them, so the agent can see what's working, spot what's wasting spend, and propose the next fix on its own.

<img src="https://mintcdn.com/adjar/AAcQKyRWyOFAa4JI/images/why-it-exists.svg?fit=max&auto=format&n=AAcQKyRWyOFAa4JI&q=85&s=a3d8b7756cedc99237587b878277384c" alt="Without Adjar, an AI agent can't work a click-based ad console. With Adjar, the agent reads and edits plain-text config and the change flows to the ad platform — Google Ads or OpenAI Ads." width="640" height="470" data-path="images/why-it-exists.svg" />

## Your account as config

Everything Adjar works with lives as plain-text files in your repo. Your campaigns, ad groups, budgets, and creatives become TOML config; your performance reports sit right beside them as Markdown. The same shape works across platforms — a `platform` field at the top of each config picks the driver — so one agent workflow spans every account. Because both config and reports are just files, any AI agent can read the full account and how it's performing, then edit the config to propose the next change.

## The loop

Adjar runs on one short loop you stay in control of.

1. **Ask.** Tell the agent what you want in plain language — "cut the keyword that's burning spend." It reads the report and the config, then drafts the change.
2. **Review.** See the full plan as a before-and-after: what changes, where, and how much. Nothing happens yet.
3. **Apply.** Approve, and Adjar pushes the change to the ad platform. Every change is logged and fully reversible.
4. **Measure.** The next report comes back and becomes your next ask.

## Who it's for

<CardGroup cols={3}>
  <Card title="Growth Marketers" icon="chart-line">
    Run paid media as a team of one. The agent reads reports and drafts the fixes; tuning that took days ships in minutes.
  </Card>

  <Card title="Marketing Agencies" icon="building">
    One workflow across every client. Launch new accounts from a proven playbook instead of rebuilding from scratch.
  </Card>

  <Card title="Heads of AI" icon="robot">
    Extend AI to the last holdout. Ads stop being the one team that can't use agents, with review and rollback built in.
  </Card>
</CardGroup>

## Why plain text changes everything

<CardGroup cols={3}>
  <Card title="Your agent can actually use it" icon="sparkles">
    Agents read and edit text reliably; a console of clicks defeats them. Plain text is what makes an agent useful at all.
  </Card>

  <Card title="It scales past clicking" icon="layer-group">
    Change 200 campaigns with one find-and-replace, not 200 rounds of clicking. Clone what works into a new market in seconds.
  </Card>

  <Card title="Every change is tracked and reversible" icon="clock-rotate-left">
    Every change is saved as a revision, so your whole account carries a version history — like code. Review the before-and-after before it ships, then revert to any earlier version if one goes wrong.
  </Card>
</CardGroup>

## Get started

<Card title="Quick Start" icon="rocket" href="/docs/quickstart">
  Get your account into Adjar in under 10 minutes.
</Card>
