What it does
The server runs as a local stdio process and makes no network calls of its own. It offers two tools. redact_suggest takes a text snippet such as a log line or prompt, detects PHI-shaped values (SSNs, MRNs, dates of birth, patient names, phone numbers and email addresses), and returns a redacted version with a confidence score for each match. scan_code walks a directory and flags lines where a sensitive identifier, such as patient, diagnosis, dob or mrn, appears on the same line as a risky sink: an LLM provider call, a console or logger call, an analytics .track( call, or an error-reporting call. It reads .ts, .tsx, .js, .jsx, .py and .go files. Matched PHI values are masked in results by default.
It is intended for healthcare codebases, where patient data is most likely to leak at the point it is interpolated into a prompt, a console.log or an analytics event, places that infrastructure scanning does not usually catch. It is a linter for that one class of mistake, not a compliance check.
Setup requires Node.js 22 or newer. The server is added to an MCP client such as Claude Code, Claude Desktop or Cursor, either from the npm release or from a local clone. Findings are returned to the client that launched the server, so if that client is a hosted assistant, the findings enter that model's context. After a rebuild, the client must be restarted or the server reconnected.
Limitations and requirements
Requires Node.js 22 or newer. It is not a HIPAA certification, compliance attestation or dataflow analyzer. Detection is purely lexical, with no parser or type information, and the pattern and identifier lists are fixed and limited to those documented. Findings go back to the launching MCP client, so hosted assistants receive them. Raw matched values are returned only when includeMatchedValues is set. A rebuild does not reach an already-running stdio process.
Safety
Our team has not reviewed this resource. Only automated rules checked it; resources with high-risk signals are not published. These are signals, not a guarantee that installing it is safe. Try it in a sandbox project first.
- Runs code or changes an assistant's behaviour
Sources
- Official MCP Registry first seen 13 Nov 2026 · last seen 13 Nov 2026 · its page on the source
io.github.Abidit/phi-guard-mcp
Local-first PHI detection for source code, LLM prompts, logs, and analytics
v0.2.0As shown on registry.modelcontextprotocol.io when we last read it