PII agent
The PII Agent helps data teams discover and monitor personally identifiable information across their data estate. It operates in two modes -- discovery and monitor creation -- and is accessible through the Operations Agent in the Monte Carlo UI or the Agent Toolkit in your IDE.
What it does
The PII Agent operates in two modes:
Discovery mode
Finds where PII already exists in your warehouse by leveraging native classification tags from your data platform:
- Snowflake — reads
SEMANTIC_CATEGORYtags applied by Snowflake's built-in classification - Databricks — reads
class.*tags from Unity Catalog
This gives you an inventory of columns that your warehouse has already identified as containing PII, without running any additional queries.
Monitor creation mode
Creates regex-based PII monitors that run in your warehouse to detect PII patterns. These monitors push metric queries to your warehouse and check match rates against configurable thresholds.
Two monitoring modes are available:
| Mode | Behavior |
|---|---|
| ALERT | Fires an incident when the PII match rate exceeds the configured threshold |
| SCAN | Silent discovery — records results without triggering incidents |
Supported PII types
PII monitors support detection of the following patterns:
| PII type | Region |
|---|---|
| Social Security Number (SSN) | US |
| Email address | Global |
| Phone number | US |
| ZIP code | US |
| State code | US |
| Credit card number | Global |
| Tax File Number (TFN) | Australia |
| Medicare number | Australia |
| Australian Business Number (ABN) | Australia |
| Phone number | Australia |
| Postcode | Australia |
Privacy-safe design
PII monitors are designed to protect sensitive data throughout the detection process:
- Monte Carlo infrastructure only sees aggregate match ratios (a number between 0.0 and 1.0) — never actual row values
- Regex evaluation happens entirely inside your warehouse
- The monitors are masking-aware — masked values (e.g.,
***-**-6789) do not trigger detections - Only unmasked PII patterns count toward the match rate
Accessing the PII Agent
Via the Operations Agent
Ask the Operations Agent in the Monte Carlo UI:
- "Find PII in my warehouse"
- "Where does my data contain email addresses?"
- "Create PII monitors for the
customersschema"
The agent follows a guided flow: scope selection, PII type discovery, mode confirmation, configuration preview (dry run), and creation.
Via the Agent Toolkit (MCP)
The pii-monitor skill is available through the Agent Toolkit in AI coding agents like Claude Code and Cursor. The skill provides access to the find_classified_assets, create_pii_monitor, and get_pii_types MCP tools.
Security & data privacy
For more detail on security and data privacy, see the AI Features and Technical Information documentation. For information about PII filtering in Monte Carlo's own data collection, see PII Filtering.