MangoseAI

MangoseAI: AI that works on your terms

MangoseAI is an assistant built into Mangose that helps with daily tasks, writing, planning, and working with information. From the start we design it so you stay in control of what goes to the model and what stays with you.

Key privacy principles

1) Your conversations stay with you

Conversation history and data are stored locally on your device, in the browser. This means:

  • we do not build a central database of your conversations,
  • you control your data without asking anyone to delete it,
  • you can clear local data at any time.

2) You decide what you share

MangoseAI works in a privacy first mode. You choose what information can be used for an answer, for example:

  • the content of your current message,
  • context from a selected space or collection,
  • specific tasks or notes,
  • or nothing.

You can use MangoseAI without sharing any data from Mangose. Then the assistant answers only based on what you type in chat.

3) EU law compliance

MangoseAI is designed to meet European Union requirements, including privacy and data minimization principles. In practice this means we collect and pass on only what is necessary to fulfill your request, and only when you allow it.

How MangoseAI works step by step

Step 1

You set the data scope

Before MangoseAI runs, you choose what context it can access. This can be minimal or fully disabled.

Step 2

Processing on Mangose servers

MangoseAI runs on Mangose servers. That is where assistant logic runs and the answer is generated based on the data you decide to share.

Step 3

Web lookup when you need it

For some tasks MangoseAI may use external tools to deliver a better answer. We always explain clearly what is used and why.

What exactly is sent outside

What may go to MangoseAI, Mangose servers:

  • your message,
  • only the context you allow,
  • technical data required to operate the service, for example request parameters.

What goes to external AI models:

  • an anonymized prompt needed to perform a specific task,
  • without your conversations, without conversation history, without data from your spaces if not needed to perform the task.
  • example: for web search we use Perplexity.

When we use external models

For specialized tasks MangoseAI sometimes uses other AI models. Then we send only an anonymized prompt we create, so the external provider cannot identify you and individual sessions are not linked. Only then, if you ask for information from the web, we use for example Perplexity search.

  • we send only an anonymized prompt needed to perform the task,
  • the external provider cannot identify you,
  • individual sessions are not linked,
  • example: for web search we use Perplexity.

In the future MangoseAI will not use other models to operate and will be self sufficient.

Your control tools

  • Data sharing choice: you decide whether and which context is used.
  • No context: you can use MangoseAI without connecting data from Mangose.
  • Cleaning: you can delete local conversation data in the browser at any time.
  • Transparency: we use clear messages about what is used and why.

Important limitations we state clearly

Each user has 500 tokens per week. The weekly token limit helps us keep the service stable and costs under control. Long term, we plan a higher limit as the service scales.

  • AI responses may contain errors or simplifications, especially on complex topics.
  • Web search depends on the availability and quality of sources.
  • If you share context, MangoseAI can use it for better answers, but you still decide the scope.

Short FAQ

Does Mangose have access to my conversation history?

Conversation history is stored locally on your device in the browser. We do not centrally store your full conversation history.

Can I use MangoseAI without sharing data from the app?

Yes. You can share no data and use only the chat.

What goes to external tools or providers?

Only an anonymized prompt needed to perform the task, without data that could identify you. Sessions are not linked. Example: for web search we use Perplexity.