What the Pagia assistant can see and do, where people confirm its work, how requests reach AI model providers, and the controls each organization holds.
Effective 14 October 2026Published by Logpagia Technologies LLC, 5900 Balcones Drive, Suite 100, Austin, Texas 78741, United States. Questions go to privacy@logpagia.com.
Last reviewed against the current source on .
Pagia is the assistant Logpagia Technologies LLC provides inside the signed-in workspace, available in the main workspace, Spaces, Contracts, CRM and the Pagia workspace. When Logpagia turns it on, it also answers in Logpagia’s own support chat and in the chat on this website, as the section on support and website chats describes. This page concerns Pagia, not Pagenia.
In the signed-in workspace, Pagia is off until an organization administrator turns it on. What a person can use then depends on their organization’s settings, their product access and permissions, and the capabilities Logpagia has released.
Pagia works within the person’s own organization and permissions. It can:
Pagia sees the person’s message, the recent conversation when history is on, the record open on the page, and the fields it reads from records the person can already open. A prompt cannot grant access to another organization’s records or widen the person’s permissions.
Pagia works only when a person sends a message. It has no background access to company data and runs no unattended work.
In action mode, creating or updating the person’s own tasks, adding subtasks and setting their own reminders run straight away, each with an Undo control. Daily work entries, personal calendar events and private documents wait for the person to confirm them.
Ask mode gives answers only and prepares no actions. In Spaces, Pagia prepares a reply draft for the person to edit and send; Pagia never sends it.
Pagia does not run payroll, make or release payments, decide on candidates or hires, sign documents, change roles or permissions, or change many records at once. These decisions stay with the people and approval steps the organization has set up.
When a person sends a message, Logpagia passes it to OpenRouter, which forwards it to the one model provider Logpagia has approved for the route, with fallback to other providers turned off. The request asks the provider to deny data collection for training. The provider’s own terms govern how long it keeps a request and where it processes it.
Before anything leaves Logpagia, it checks the message and selected records and blocks detected credentials, bank and payroll details, government identifiers and health information. Logpagia stores permitted conversations in the workspace and keeps metadata-only evidence of each provider request, with no message text. Logpagia does not use workspace records to train AI models.
Logpagia decides whether Pagia answers in its support chat and in the website chat, separately from each organization’s Pagia settings. In the support chat, Pagia answers questions about Logpagia’s products from Logpagia’s product guide; in the website chat, it also answers questions about plans from Logpagia’s plan catalog. It sees the conversation only, reads no account or organization data and takes no actions. Its first reply says that it is an AI assistant, and anyone can ask for a person at any time.
Every other request, including a demo, goes to the Logpagia team as a ticket, and a member of the team replies in the chat or by email. Website chats have their own daily limit; once it is reached, website chats go straight to the team. Visitors give a name and an email before chatting, and Logpagia staff act on an account request only after the person signs in and opens Help and support.
With web search on, Pagia writes a short standalone query of at most 300 characters and sends only that query to the same provider. The conversation, workspace records, remembered facts and attachments stay out of every search, and each reply lists the pages it read as sources.
With memory on, a person can keep up to 20 short facts about themselves, such as how they like answers. Facts about other people, credentials, identifiers, health, pay and banking details are refused. Each person can edit or delete their facts at any time, administrators never see them, and deleted facts are removed permanently after 30 days.
An administrator with the Pagia management permission turns Pagia on by accepting the Pagia policy once, and can turn it off at any time. That one acceptance covers recent conversation history, web search and memory, each has its own switch, and every change is recorded in the organization’s audit log.
Roles decide which members can use Pagia and its features. Each member can turn their own memory off, and Logpagia can stop all external processing platform-wide at once.
Pagia output can be wrong, incomplete, out of date or a poor fit for the request. Review the source, context and proposed change before relying on an answer or confirming a draft, and confirm only what you understand.
Pagia output is not professional advice. Consequential employment, financial, legal, compliance, safety and access decisions stay with a qualified person.
Users keep secrets and unnecessary sensitive information out of prompts, respect their organization’s permissions, and check generated content for text designed to mislead the assistant or extract information. Pagia is for the organization’s legitimate work; using it to bypass a control or to harm another person is prohibited.
Questions about this policy go to privacy@logpagia.com. The in-product Pagia policy that an administrator accepts carries its own version number; when it changes what leaves the organization, Pagia pauses until an administrator accepts the new version.