> ## 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.

# Audit trail

> What Surfaice logs about agent runs, who can see it, and how to use it in reviews.

Enterprise teams need to answer: **who asked the agent to do what, using which systems, and what was produced?**

**Hugo** is Surfaice’s in-app AI agent — “agent runs” and “Hugo runs” mean the same thing. Surfaice maintains an audit trail of that activity for workspace administrators and compliance reviewers.

## What is logged

| Event                           | Typical fields recorded                                                                   |
| ------------------------------- | ----------------------------------------------------------------------------------------- |
| User message / skill invocation | User identity, workspace, timestamp, channel                                              |
| Tool and connector use          | Which integration was called (for example Lucernex read, SharePoint file, mailbox search) |
| Agent output                    | Artifacts generated, citations, saved workspace objects                                   |
| Errors                          | Failed tool calls (for support diagnosis)                                                 |

Exact retention period and export format depend on your customer agreement. Request details in the NDA security package via [security@surfaice.pro](mailto:security@surfaice.pro).

## What is not a substitute for audit logs

* **PostHog / product analytics** — funnel and usage metadata, not full message bodies
* **Chat UI scrollback** — convenient for users, not the compliance system of record
* **Source systems** — Lucernex and SharePoint retain their own access logs

## Who can access audit data

* **Workspace administrators** — operational review and support
* **Your IT / security team** — via agreement-defined export or questionnaire process
* **Surfaice support** — only when engaged for an incident, under contract

End users should not expect to browse every other user's prompts unless your admin has granted that capability.

## Using audit trails in practice

**During rollout:** sample 10–20 agent runs per week. Confirm connectors match approved scope and no unexpected mailbox reads occur.

**For Vanta / SOC 2:** point reviewers to [Security overview](/security), [Trust FAQ](/concepts/trust-faq), and the NDA Application Architecture Diagram for logging boundaries.

## Related

* [Trust FAQ](/concepts/trust-faq)
* [Data handling](/concepts/data-handling)
* [Permissions](/concepts/permissions)