Control & security

    You decide what the agent can do

    The agent has no free-roaming access to your systems. It works exclusively with the capabilities you've enabled in its settings.

    • Only assigned functions
    • Only selected files and folders
    • You manage the knowledge base
    • Full visibility into every conversation

    Scope of access

    Your settingsfunctions, knowledge, files
    Agentworks within those settings
    Only allowed operationsnothing beyond what's assigned
    Overview

    What you're in control of

    Assigned functions

    The agent can only use the functions you've assigned it. Nothing else.

    Selected files

    You give the agent access to specific files or entire folders – never the whole library.

    Conversation history

    You can open any conversation at any time, including the functions used and its output.

    Account access

    Login via email or Google, with support for two-factor authentication and trusted devices.

    Data protection

    Agents are built with GDPR in mind

    Our agents only work with the data they truly need to complete a task. System functions that return order and product data are deliberately free of personal customer data – no names, addresses, emails or phone numbers.

    No customer names

    Order data never includes customer names – the agent never sees or processes them.

    No addresses

    Delivery and billing addresses are never passed to the agent.

    No emails

    Customer email addresses are never returned by system functions.

    No phone numbers

    Customer phone numbers are never part of the transferred data.

    Data minimization

    The agent only receives the data strictly necessary for the task – nothing more.

    GDPR compliant

    Personal data processing is designed to comply with applicable legislation.

    Personal data protection

    Personal data never reaches the AI in its original form

    When a customer mentions sensitive personal data (email, phone number, IBAN, etc.) in an email or message, our system encrypts it into an unreadable form before it is sent to the AI. The AI only works with the encrypted value, and if it uses it in a reply, the system decrypts it back to its original, readable form once it returns. Sensitive data therefore never reaches the AI.

    Two levels of protection

    In the agent's settings you choose how wide a range of personal data should be encrypted.

    Basic protection

    Encrypts common sensitive data – email addresses, phone numbers, IBANs, card numbers and IP addresses. Suitable for most e-shops and recommended as the default setting.

    Maximum protection

    Encrypts everything from basic protection plus further identifiers – numeric IDs, sensitive tokens, passwords and more. For situations where you want the strictest coverage.

    How it works

    Encryption example

    A customer writes an email with contact details. Our system recognises the email and phone number, encrypts them, and the AI only receives the encrypted form.
    1Data arrivesemail and phone number from the customer
    Encryptionthe system converts the data into a cipher
    AI processingthe AI works only with the cipher
    Decryptionthe reply is returned to a readable form

    Before and after being sent to the AI

    Customer email

    jan.novak@example.com↓{{email:5f1v33e51500fec}}

    Phone number

    +421 901 234 567↓{{phone:e5d68d60f2444d92}}

    The AI only sees the encrypted values on the right. When it uses them in a reply, our system decrypts them back to their original form once they return, so the customer and you see a readable value – and sensitive data never reaches the AI.

    Settings

    Settings available for every agent

    For each agent you configure both its focus and the scope of what it can do.
    • agent type and communication language
    • AI model based on task complexity and cost
    • additional instructions for the agent's behavior
    • list of allowed functions
    • access to the knowledge base
    • access to selected files and folders
    • connecting a mailbox or LiveAgent
    • automated tasks and their schedule
    Transparency

    Full visibility into every task

    For every conversation you can see the request, the functions used, the processing steps, and the final output. Any generated documents can be downloaded.
    1Requestfrom a user or from a schedule
    Functions usedvisible in the conversation
    Processing stepsstep by step
    Output and filesready to download

    Start with an agent that has clear boundaries

    Give the agent exactly the functions, knowledge and files it needs to do its job.