> ## Documentation Index
> Fetch the complete documentation index at: https://docs.surfaice.pro/llms.txt
> Use this file to discover all available pages before exploring further.

# Data handling

> What Surfaice persists, what stays ephemeral, and how model providers are used.

## Persistence model

Surfaice assembles context at run time for a task. It does **not** keep a general-purpose copy of your document repositories.

**Skills** are named workflows the agent (Hugo) can run; **workspace configuration** includes which skills and instructions your company enabled. New here? See [What is Surfaice?](/concepts/what-is-surfaice).

### What typically persists

* Accounts and user identity
* Encrypted connection tokens
* Workspace configuration, skills, and instructions
* [Audit and agent-run logs](/concepts/audit-trail)
* Content a user deliberately saves (for example an accepted lease abstract moved into workspace memory)

### What Surfaice is not storing as a product default

* Bulk copies of mailboxes
* A mirror or sync of SharePoint or your DMS
* A general-purpose lake of customer documents
* Customer data used to train any model (ours or a vendor's)

## Hosting and isolation

* Each customer gets a dedicated, isolated environment with its own database
* Hosted on Google Cloud Platform in US regions
* Encrypted at rest and in transit

## Model providers

Inference may use commercial agreements with providers such as Microsoft Azure OpenAI, Google Gemini, and Anthropic, under zero-retention / no-training contractual terms for customer content as described on our [security page](https://www.surfaice.pro/security).

## Authoritative security summary

For subprocessors, compliance status, and the NDA security package, use:

* Public summary: [surfaice.pro/security](https://www.surfaice.pro/security)
* Docs page: [Security overview](/security)

Do not treat marketing blog posts as the source of truth for security questionnaires.