AI help assistant
What is the AI help assistant?
The help assistant is an optional chat feature built into the EmailConnect dashboard. It can answer questions about configuration, webhooks, domains, aliases, and other platform features — drawing on our knowledge base articles.
You can open it with the ? icon in the navigation bar, or press Ctrl+K / Cmd+K.
What data does the assistant receive?
When you send a question, the following context is included so the assistant can give relevant answers:
| Data sent | Example | Why |
|---|---|---|
| Your question | "How do I verify a domain?" | To answer it |
| Plan identifier | business |
To tailor answers to your plan's features |
| Current screen | aliases |
To suggest relevant articles |
| Resource counts | 2 domains, 5 aliases | To give contextual guidance |
What is never sent:
- Email content or message bodies
- Email addresses, sender/recipient information
- Webhook URLs or secrets
- API keys or credentials
- IP addresses or browser fingerprints
- Any personally identifiable information (PII)
Data handling: streaming only, never stored
All conversations are streamed in real-time and are never persisted:
- Questions and answers exist only for the duration of the HTTP request
- No chat history is stored on our servers or by the AI provider
- Closing the help panel discards the conversation entirely
- There is no training on your data — your questions are not used to improve the model
Fully EU-operated infrastructure
Every component of the AI assistant runs within the European Union:
| Component | Provider | Location |
|---|---|---|
| Application backend | Hetzner | Germany |
| Knowledge base service | Hetzner | Germany |
| AI model (Mistral) | Scaleway | France |
| Vector database | Hetzner | Germany |
We specifically use Scaleway's Mistral API rather than Mistral's own API endpoint, because Mistral's primary API runs on Microsoft Azure infrastructure. Scaleway's wrapper guarantees the model runs on EU-owned infrastructure in France, with no data leaving the EU.
Rate limits
To prevent abuse, the assistant is limited to 20 questions per minute per user. This resets automatically.
Opting out
The assistant is entirely optional. If you never open the help panel, no data is ever sent to the AI service. There is no background processing, telemetry, or passive data collection related to the assistant.