Introduction

Note

Proof of concept. This extension explores a concrete question: is agent-like behavior possible within the TYPO3 backend? It is not intended to answer whether this is the right architectural approach — the space is moving fast, and the tradeoffs between MCP, tool-calling, browser-side agents, and custom integrations are far from settled. The goal here is to show that it works, and to invite feedback from anyone thinking about the same problem. If you have thoughts, open an issue.

What does it do?

AI Chat adds a backend module to TYPO3 that lets administrators and editors interact with an AI assistant directly from the TYPO3 backend. The module is available under Admin Tools > AI Chat.

Using the tools nr-llm registers, the assistant can read and modify TYPO3 content -- pages, content elements, records -- from natural language instructions. All processing happens server-side via CLI commands, keeping the web server responsive.

AI agent creating a page, adding content, and optimizing SEO in TYPO3 — The AI agent creates a page, adds content, optimizes SEO fields, and evaluates the result — all via natural language in the TYPO3 backend.

Key features

Integrated chat module

A dedicated backend module under Admin Tools with a modern chat interface. Send messages, view responses, and manage conversations without leaving TYPO3.

Tools from nr-llm

The chat runs on nr-llm's agent runtime and tool registry: builtin read and write tools, plus any MCP server configured in nr-llm's MCP Servers module.

Conversation history

Conversations are persisted in the database. Resume previous chats, pin important ones, or let the system auto-archive inactive conversations.

Floating chat panel

A toolbar button opens a resizable bottom panel that stays visible across all module navigation. Chat while working in the page tree without switching context.

File attachments

Attach PDF, DOCX, TXT, and XLSX files to your messages. Text is extracted server-side when needed, so all formats work regardless of the LLM provider. Vision-capable providers also accept images -- which ones depends on the provider (PNG, JPEG, GIF and WebP everywhere, HEIC and HEIF on Gemini).

Markdown rendering

AI responses are rendered as rich Markdown -- headings, lists, code blocks, and tables -- using marked.js with DOMPurify for XSS safety.

Secure by design

Access is restricted to configured backend user groups. Messages are length-limited, concurrent conversations are capped, and API keys are never exposed to the browser.

Example interactions

Once a provider is configured in nr-llm, you can ask the assistant to perform tasks like:

"Show me all pages under the homepage"
"Create a new text content element on page 42 with the heading 'Welcome'"
"What content elements exist on page 15?"
"Move the news page to be a subpage of 'About Us'"
"List all hidden pages in the site"

What it can actually do depends on which tools nr-llm makes available: its builtin set reads the installation and makes bounded editorial writes (new pages and content elements arrive hidden, as drafts), and an MCP server registered under AI > Operation > MCP Servers adds whatever tools that server offers. With every tool switched off, the chat still works as a general-purpose assistant on the configured LLM provider, but cannot see or change TYPO3 content.

Acknowledgments

This extension builds on the work of others:

nr-llm: The Netresearch LLM abstraction layer for TYPO3 that provides provider-agnostic access to language models.
nr-vault: Secure credential storage for TYPO3, used to protect API keys for LLM providers.