> ## Documentation Index
> Fetch the complete documentation index at: https://docs.get3rd.com/llms.txt
> Use this file to discover all available pages before exploring further.
# 3RD Agent
> Ask 3RD anything about your visibility — and have it do the work. The agent reads your data, explains what it finds, and makes changes you approve.
The **3RD Agent** is the assistant built into every page. Ask it a question in plain language and it reads your actual data to answer — then offers to do something about it.
It is not a search box over the docs. It runs the same queries the dashboard does, and it can change your setup: add prompts, add a market, build a playbook, create a report, draft a page. Every change is shown to you first as a card you can edit or reject.
## Where to find it
* **In the sidebar** — **3RD Agent** opens the full page, with your conversation history down the left.
* **On every page** — the **Agent** tab in the top-right corner opens a panel beside what you are looking at. Ask about the page you are on without losing your place.
The panel knows which page you opened it from, so its suggestions change with you: on **Sentiment** it offers to explain what people are saying, on **Reports** it offers to build one.
## Three things worth knowing
Every answer lists what it read to get there, and the figures come from the tool output rather than being retyped — so it cannot misquote your own numbers.
Any change arrives as a card showing the exact fields. Edit them, approve, or decline. Nothing is written until you press approve.
Hover a chart or a card and choose **Ask the agent** to ask about that specific thing, with its numbers already attached.
## What it can do
The agent has **45 skills**. Open **Explore skills** in the agent's sidebar to browse them, or just ask — you do not have to pick one.
### Where you stand
| Skill | What it answers |
| - | - |
| How are we doing | The headline figures, and whether they are good. |
| Why it moved | What caused a rise or a fall, not just that it happened. |
| What each model says | Where you stand engine by engine, and what that means. |
| What did we lose | Which prompts you dropped out of, and who took the place. |
| Why is this empty | The reason a number is zero, and the specific fix. |
### What's being said about you
| Skill | What it answers |
| - | - |
| Is what they say true | Checks the claims the models make about you against the facts. |
| Perception gap | How the models describe you, against how you want to be described. |
| Who gets cited instead | The sources answering your questions, and why they win. |
| Are we named consistently | Whether you are described the same way everywhere models look. |
| What people actually ask | The real questions people put to an assistant about a business like yours. |
| Can I trust this number | What a figure counts, and where it stops being reliable. |
| Are our topics still right | Whether prompts are filed under topics that still describe them. |
### Your pages, products and ads
| Skill | What it answers |
| - | - |
| Why isn't this page cited | What the cited pages have that yours does not. |
| How good is this page | Readability, structure and what would improve it. |
| Can the models reach us | Whether the AI crawlers can fetch and read your pages. |
| How my ads are doing | Campaign performance, spend and what is working. |
| What rivals are advertising | The ads your competitors are running right now. |
| Which products show up | Where your products appear when people shop by asking. |
### Decide what to do
| Skill | What it answers |
| - | - |
| What to write next | The content gaps that cost you the most, ranked. |
| Prompts worth adding | Closes the gaps in what you are tracking. |
| Rivals you are missing | Brands appearing beside you that you do not track yet. |
| Build a playbook | Agrees what to influence, then plans the actions. |
| Argue against my plan | The case against what you are about to commit to. |
| Is my plan still right | Checks the playbook against what the data says now. |
| Which pages to watch | Picks the pages worth tracking, and says why each one. |
| Which products to watch | Chooses the products worth running prompts for. |
| Where to expand | A country worth watching next, argued from what the answers already show. |
### Make something
| Skill | What it does |
| - | - |
| Build a report | Picks the sections, the cadence and who receives it. |
| Turn a gap into a brief | A writing brief from a gap worth closing. |
| Write or rewrite a page | Drafts the page a recommendation asks for. |
| Draft ad creative | Ad copy to review before it goes anywhere near live. |
| Turn this into a task | Sends an answer straight to your playbook as work to do. |
| Make a chart and keep it | Turns an answer into a chart, and keeps it on a report that rebuilds it. |
### Set things up
| Skill | What it does |
| - | - |
| Add prompts | Covers a topic you are not tracking yet. |
| Add a market | Tracks a new country, in its own language. |
| Add a competitor | Tracks a rival you are being compared against. |
| Create a brand | Sets up a new brand to monitor. |
| Change AI models | Chooses which models answer your prompts. |
| Fill in the profile | Asks a few questions and fills the gaps in your profile. |
| Set our objectives | What you are trying to achieve — it steers every recommendation. |
| Upload a list of pages | Adds many pages at once instead of one at a time. |
### Tidy up
| Skill | What it does |
| - | - |
| Tidy up prompts | Edits or removes prompts in bulk, from a list you paste. |
| Fix the wrong language | Finds prompts left running in a language the market no longer uses. |
| Re-pick tracked pages | Changes a page selection that got frozen. |
| Fix a product that isn't ours | Claims a product, or corrects one attributed to the wrong maker. |
## Getting a useful answer
"Why did we drop in Germany last week?" works better than a keyword. The agent picks the right data itself.
The steps under each answer say which data it read. If it looked at the wrong window or the wrong market, say so and it will redo it.
Answers end with one or two suggestions — turn this into a task, add the prompts that close this gap, build a report. They are derived from what it actually did, not a fixed menu.
## Answers you can keep
* **Charts** can be downloaded as an image, a PDF, or the numbers behind them as CSV.
* **Tables** come with the full data as CSV, even when the answer only shows the first few rows.
* **Anything it creates** — a report, a task, a page — comes back with a link straight to it.
The agent works on the brand you currently have selected, and respects the same market, date and model filters as the page you opened it from. Switch brand and the conversation starts fresh.
## Limits worth knowing
* It **cannot** change billing, delete a brand, or send anything outside 3RD without you approving it first.
* It answers from the data 3RD has collected. If a market has not run yet, it will tell you that rather than estimate.
* Changes that cost money — adding a market beyond your plan, for example — are refused with an explanation rather than made quietly.
## Related
The plan the agent builds and keeps current.
Where the tasks it creates land.
Drafting a full page from a content task.
The same capabilities from your own tools.
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.