Start your 7-day free trial, card not charged until it ends
Browse all topics

The in-app assistant

Apps & integrations · Last updated 11 September 2026

The assistant is a chat window built into the dashboard. You ask it a question about your sites in plain English, it looks up the real data behind the answer, and it can carry out a small set of jobs for you once you approve them.

What it is

A conversation panel inside TLDTrack. It reads your own account data through the same read-only tools the API uses, answers questions about DNS, email, uptime, SSL, scans and alerts, and can propose a small number of actions. Every action stops at a confirmation card first, so nothing is changed or run until you press Confirm.
Read more
The assistant is scoped to one organisation. It can see the sites, checks, alerts and scan results belonging to the account you are signed in to, and nothing else. It cannot browse the wider internet on its own. When it needs a value it calls a tool that queries your data and quotes what came back. Conversations are saved, so you can reopen an earlier thread from the list on the left and continue it.

Why it matters

Most of what TLDTrack knows is spread across tabs. Finding out why email is failing, or which certificate expires next, means knowing which screen to open. Asking a question is faster, and the answer comes back with the real record or date attached rather than a general explanation.
Read more
The difference between a chatbot and a useful assistant is whether it looks things up. This one is told to check before it answers, to quote the actual value it found, and to say plainly when a tool returned nothing rather than filling the gap with a guess.

This is not the same as API and AI access

There are two separate things with similar names. API and AI access is about letting an outside tool, such as Claude or ChatGPT, read your TLDTrack data through an API key or a connector. This article is about the assistant that lives inside the dashboard. You do not need an API key to use it.

Where to find it

There are two ways in, and both use the same assistant:

  • The full page. Open Assistant from the sidebar. You get the chat on the right and your saved conversations on the left.
  • The side panel. The speech-bubble button in the top bar slides the assistant out over whatever page you are on, so you can keep the data in view while you ask about it.

If you open the panel while you are on a site's detail page, the conversation starts focused on that site. The header above the chat tells you which, or says "All your sites" when nothing is in focus.

Simple and Advanced

A toggle at the top of the chat switches between two ways of answering. The setting is saved against the conversation, so a thread keeps the mode you chose for it.

  • Simple. Plain English, one question at a time, jargon explained the first time it appears. It walks you through a problem rather than listing everything it found. When it spots something it can fix, it offers the fix.
  • Advanced. Terse and technical. It shows the raw values it looked up, the actual DNS records, the SPF and DMARC strings, the scores and dates, and answers the question directly without the hand-holding.

What it can look up

Before answering, the assistant can query your own data: the sites you monitor, their DNS and email records, uptime and SSL state, blacklist and malware status, email deliverability results, security, compliance and brand results, content and visual changes, alerts, clients, tasks and time. It can also report your current monitoring-credit balance.

Turning a module off in Settings also takes its tools away from the assistant. If you have switched off compliance, for example, the assistant has no compliance tool to call and will say so rather than inventing a result.

What it can actually do

Beyond answering, the assistant can propose six jobs. Each one arrives as a confirmation card showing what it is about to do and, where the job costs credits, how many. Nothing happens until you press Confirm, and you can press Decline instead.

  • Run an email deliverability test on one of your sites.
  • Run a scan: security, SEO, accessibility or compliance.
  • Acknowledge an alert, so it stops counting as needing attention.
  • Change which DNS record types raise a change alert for a site.
  • Create a page task or note.
  • Create a service monitor, meaning an HTTP endpoint or TCP port health check.

Confirmation is the boundary

The assistant cannot run a scan, change a setting or spend a credit on its own. Those tools are separated from the read-only ones in code, and the only path to running one is a card you approve. If a reply claims something was done, check that a confirmation card appeared and you accepted it.

What it will not do for you

It cannot set up a content or visual monitor, because that means choosing the exact value or region on a page, which only you can point at. Rather than refusing, it explains what the monitor does and links you to the setup page for that site. The same applies more broadly: when a job is outside its tools, it is told to send you to the exact screen that can do it, as a link, with the short version of the steps.

Credits and limits

  • It uses monitoring credits. The assistant draws on the same shared credit pool as scans. If the balance is empty it stops and tells you to top up on the billing page instead of answering.
  • Actions cost separately. A confirmation card that will spend credits says how many before you accept it.
  • Thirty messages an hour, per person. Past that it asks you to try again shortly. The count is per user, not per organisation.
  • Two thousand characters per message. The message box will not take more than that.

See Scan credits for how the pool works and Credit packs for topping it up.

Conversations

Each thread is saved and titled from your first message. The list on the left of the full page holds your previous conversations; select one to carry on where you left off, or start a fresh one with New conversation. Threads belong to you, not to the whole team.

After each reply the assistant offers up to three short follow-up suggestions as buttons. They are shortcuts, not a menu: you can always type something else.

Getting a useful answer

  1. 1
    Open the assistant from the sidebar, or from the speech-bubble button in the top bar while you are looking at a site.
  2. 2
    Pick Simple or Advanced depending on how much detail you want back.
  3. 3
    Name the site in your first message, unless you opened the panel from that site's page. After that it assumes follow-up questions are about the same site until you name a different one.
  4. 4
    Ask the real question. "Why are my emails going to spam on example.com" gets further than "explain SPF", because the first one makes it go and read your records.
  5. 5
    If it proposes an action, read the confirmation card, check the credit cost, then confirm or decline.

Where to find this

Good to know

  • It only ever sees the organisation you are signed in to.
  • Text it reads from your web pages is treated as data, never as instructions, so a page cannot tell the assistant what to do.
  • Domains are always referred to by hostname. It will not show you internal ids.
  • Switching a module off in Settings removes the matching tools from the assistant as well as the tabs.
  • It is a technical advisor for your sites. Ask it something unrelated and it will steer back.
Available on:StarterProAgency

Was this guide helpful?

If anything is unclear or missing, tell us and we will improve it.

Contact support