Skip to main content

Know what is checked

Default checks cover email, SSN, credit cards, labeled US routing numbers and NPIs, and supported credentials. Phone, date, ZIP code, and IP address checks are available but off by default. Plain-text names and street addresses are not detected. An explicit entity_types selection replaces the defaults. Include every type you want to check or transform. A clean scan means no selected detectors matched; it does not prove anonymity. Supported inputs are UTF-8 text files up to 10 MB, including CSV, TSV, JSON, logs, ENV, and SQL. CSV/TSV headers are preserved and not scanned by default; use has_header=false when the first record must be checked. PDF, DOCX, XLSX, images, and archives are not parsed. A refused scan means the file was not checked.

Understand the privacy boundary

The running server processes file contents locally and makes no network requests. Tool responses omit file contents and matched values, including when an operation fails. Paths and metadata still enter the agent’s context and may reach its model provider. Allowed roots restrict DataFog’s access. They do not restrict the agent’s other tools, and asking it to scan first does not enforce that workflow. The agent can still open a file directly. Other MCP clients and Windows have not been validated.

Further reading