Chat widget anywhere
Drop the Chatbot content element on any page and configure it in the FlexForm. Intro message, placeholder and labels are translatable.
TYPO3 extension · undkonsorten/easychat
EasyChat puts an AI assistant on your TYPO3 website without a third-party chat platform. Conversations are stored only in your TYPO3 database. All you need is TYPO3 and an LLM endpoint.
composer require undkonsorten/easychat
Live demo
Styled or unstyled, with or without the built-in consent modal. Every variant runs on this site.
Styled widget (bottom right) with the built-in consent modal.
Styled widget (bottom right), no consent required.
Default Deep Chat look with the built-in consent modal.
Default Deep Chat look, no consent required.
Why EasyChat
Features
Content element, records, Reactions, Scheduler and Fluid. Editors and integrators work with what they already know.
Drop the Chatbot content element on any page and configure it in the FlexForm. Intro message, placeholder and labels are translatable.
Three consent modes, a random session ID and the messages. Nothing more. A Scheduler task deletes old chats automatically.
Any OpenAI-compatible endpoint via Symfony AI. Switch provider by changing one record. The model list loads live from the provider.
Retrieval-augmented generation with EXT:index and Qdrant. Pages, content records and files become the bot's knowledge. Only public content is embedded.
Browse sessions, read full conversations, export them as CSV or delete them. Editors can remove a single session on request.
Run it with German AI hosting such as mittwald and gpt-oss-120b, with Mistral, or with a fully self-hosted model.
Setup
Optional: Scheduler task for cleanup · Styled template via static TypoScript · RAG with EXT:index and Qdrant. Full documentation
composer require undkonsorten/easychat
Create an EasyChat configuration record with URL, API key, model and system prompt.
Create a TYPO3 Reaction of type EasyChat that points to the configuration.
Add the Chatbot content element and paste the reaction URL and secret. Done.
Proven in practice
Reviewed and approved by the data protection officers of PricewaterhouseCoopers (PwC).
EasyChat is running in production and answers real visitors' questions every day.
Support by undkonsorten
EasyChat is open source and stays that way. If you want a partner for the project, the team that builds the extension helps you plan and run it.
Which LLM fits your content and your privacy requirements, and where it should run: European AI hosting or your own server.
Installation, Reaction and configuration, and a chat widget styled to your corporate design.
Indexing your pages and files for RAG, and a system prompt that keeps the bot on topic and in your tone.
Roadmap