@phi-longevity/mcp-server

Name: phi mcp server
Author: philongevity

A Model Context Protocol (MCP) server that lets AI agents call Phi Longevity's PRISM clinical recommendation engine on synthetic biomarker panels — guideline-cited, evidence-tiered recommendations an agent can use when researching options for the person it's helping.

Built for chronic-condition work in particular: type-2 diabetes, lupus, and cancer survivorship each have their own condition lens, because the same panel means different things depending on what someone is managing.

🛡️ Synthetic / de-identified data only. Do not submit protected health information (PHI). The server is stateless and stores nothing. Real health files are analyzed only inside the authenticated Phi Longevity app, after consent, by the account owner.

Tools

Two surfaces, deliberately different. The npm package (stdio) has 4 tools; the remote endpoint has 6. The two report-generating tools are remote-only.

Tool	npm	remote	What it does
`sample_prism_report`	—	✅	Start here. Returns a complete, real PRISM report for a sample chronic-condition persona in one instant call. Free, no signup, no input required. The fastest way to see what the engine actually produces.
`quick_check`	✅	✅	A fast read on a handful of values — the low-friction entry point before a full analysis.
`analyze_biomarkers`	✅	✅	Analyze a synthetic panel → tiered, guideline-cited recommendations, with a condition lens.
`list_supported_biomarkers`	✅	✅	51 scored biomarkers + units + reference ranges, by clinical pillar.
`get_methodology`	✅	✅	How the Phi Score works (5 pillars + weights) + link to the full methodology.
`full_prism_report`	—	✅	A full report for a panel you supply. Paid — priced per call via x402; terms are returned with the offer.

All tools except full_prism_report are free.

Quick start — zero config

Remote (zero-install) — recommended

Point any MCP client that supports Streamable HTTP at:

https://philongevity.com/mcp

Nothing to install, no key to manage, all 6 tools. Try sample_prism_report first — it needs no arguments and returns a complete report immediately.

Claude Desktop / any MCP client (stdio)

{
  "mcpServers": {
    "phi-longevity": {
      "command": "npx",
      "args": ["-y", "@phi-longevity/mcp-server"]
    }
  }
}

A published, rate-limited access key is built in. If you have a dedicated key, set PHI_MCP_KEY to override it. This path exposes the 4 tools marked ✅ above.

Run locally

npm install && npm run build
node dist/index.js

Example

// sample_prism_report — no arguments, returns a complete sample report
{}

// analyze_biomarkers input (synthetic values only)
{
  "biomarkers": { "HbA1c": 6.4, "LDL-C": 145, "Triglycerides": 190 },
  "conditionFocus": "type2_diabetes"   // or general_wellness | lupus | cancer_survivorship
}
// → tiered, guideline-cited recommendations + a `full_report` block
//   pointing the user to a complete PRISM report at philongevity.com

For the person you're helping

The tools above analyze a handful of values at a time. A full PRISM report consolidates all of a person's lab reports, wearable data, and clinical notes into one integrated picture with a personal health score and progress over time — which is the actual problem for someone managing a chronic condition across several clinicians who don't talk to each other. Agent docs + signup: https://philongevity.com/for-agents

Environment

Var	Required	Notes
`PHI_MCP_KEY`	no	Overrides the built-in published key. Rate-limited + revocable.
`PHI_ENGINE_URL`	no	Override the default Phi MCP gateway URL (rarely needed).

HIPAA / privacy

Synthetic-only at the protocol boundary; stateless; aggregate-only telemetry (counts/timing to stderr, never values or recommendation text); zero access to Firestore / the HIPAA datastore / user accounts. GDPR: processes no personal data (synthetic only) → minimal exposure.