Ask your TYPO3 anything: 15 questions answered in seconds

TYPO3 content questions are an easy first step with TYPO3 MCP. Ask what a page says, which pages are hidden or what a tariff costs, and your AI assistant reads the answer straight from your backend. Below are fifteen questions in plain words, each answered on our demo install with one read-only tool call. For each, you see the tool the assistant used and the answer it gave. MCP, the Model Context Protocol, lets an assistant such as Claude work inside TYPO3 through tools that AI Foundation provides for free. All fifteen took under a minute together, and none of them changed anything.

Most questions people ask about a TYPO3 site fall into three groups:

• Facts from your content. Prices, opening hours, warranty terms, the values in a table. The assistant reads the content elements of the page, tables included.
• The state of your site. Hidden pages, scheduled pages, languages, redirects. The assistant reads the page properties and the site configuration.
• Housekeeping. Empty pages, old phone numbers, switched-off redirects, a statement nobody has reviewed for years. The assistant searches or counts across the whole site.

You ask in your own words. The assistant picks the tool, reads, and tells you where the answer comes from, so you can open the page in the backend and check.

We asked these fifteen questions on 6 October 2026, as an administrator working on Live. The demo runs two fictional brands: Mustermann Outdoor, which sells e-bikes, tents and jackets, and Stadtwerke Musterstadt, a utility for electricity, water and buses.

Every question used one core tool of AI Foundation, which is free. The answers are summed up here; your assistant words them in its own way.

Behind each answer sit three steps, and all of them only read:

1. Find the place. The assistant looks up the page, either by title or from the page tree. If you name the page or the site, it gets there faster.
2. Read what is there. It lists the content elements of the page, or the page properties, or the records of one table. Tables in the text are read cell by cell, so it can quote one row.
3. Answer with the source. A good answer names the page or record. If it does not, ask: Which page did that come from?

For questions across the whole site, such as how often an old hotline still appears, the assistant counts with record_count. Counting is fast and returns no text, so for the list itself you ask a second question. News sits in an extension table; the assistant reaches it because that table is enabled through Table Discovery. It reaches only the tables you enable.

The tariff price. Asked what Musterstadt Green costs, the assistant opened Electricity tariffs, read the table in the element Prices from 1 January 2026 and quoted one row: EUR 12.90 base price per month and 33.9 cent per kWh, including VAT. It also said where the prices came from, so a colleague can check the same cell.

The empty page. Asked which page has no content, the assistant listed the content elements of Events and found none, in English or German. That page is live and in the menu, so a visitor sees an empty page. Filling or hiding it is an edit, and episode 3 shows everyday edits from chat.

The old hotline. Asked how many content elements still show 0800 111 222, the assistant counted 17. It did not list or change them. That is the right order: you know the size of the job before you start it. Changing a phone number everywhere is the second post of episode 3.

When an answer is missing or wrong

The assistant answers from what it can read. Three things to know:

• Ask for the source. If an answer surprises you, ask which page or record it came from, and open it in the backend.
• Some fields are not returned. On our demo, Who last changed the E-Bikes page? got no answer, because the page tool did not return the change date. The assistant said so instead of guessing.
• Hidden and scheduled pages count too. The assistant reads the backend, not the website, so it sees pages a visitor does not. Question 1 is useful for exactly that reason.

The video for this episode asks two of these questions in Claude on our demo install, the tariff price and the hidden pages, and shows the tariff answer next to the page it came from. Shown in Claude; other MCP assistants work too. Nothing is changed.

Every question here uses AI Foundation's core tools, which are free and open source, no licence key, no registration. AI Assistant, AI Accessibility, AI Chatbot and AI Search add their own MCP tools for writing, translation, alt text, chatbots and search. Each needs its own licence. Together with AI Foundation they bring 100+ MCP tools.

What your assistant reads, it sends to its vendor, as with any chat. Keep that in mind before you ask about pages that are not public yet.

The second post of this episode asks the same kind of questions about SEO. It covers missing descriptions, duplicate titles, an accidental noindex and the canonicals. Next week, episode 5 brings your first prompts in the Playground and workspaces for changes that need a review first. New to the series? Start with TYPO3 MCP use cases, the map of jobs you can hand over, and everyday edits from chat for answers that need a change.

The series comes from T3Planet by NITSAN Technologies, a TYPO3 Association Gold Member with more than 15 years of TYPO3 work. Its episodes are written by Jürgen, our AI Consultant for TYPO3. Every episode runs its job on our demo install before it is published.

All episodes of MCP in Practice for TYPO3