oak-longevity-mcp-server

An MCP server for longevity & metabolic medicine — a medication catalog, evidence-based dosing protocols, contraindication screening, drug-interaction checks, required baseline labs, ongoing monitoring plans, FDA/compounding regulatory status, and patient-intake pathway suggestions across 35 compounds.

Built for Oak Longevity Institute by Keith Schmidt, MD — a telemedicine longevity practice in Illinois. This server makes the practice's clinical reference data available to any MCP client (Claude Desktop, Claude Code, or your own agent), and is structured for a free/premium monetization model.

⚠️ Clinical decision-support, not medical advice. All output must be reviewed by a licensed clinician. Many longevity compounds here are used off-label, are compounded, or are investigational/not FDA-approved. Dosing, contraindication, interaction, lab, and regulatory data change frequently — always verify against current primary literature, FDA/DEA resources, and your state board of pharmacy.

What it does

The eight categories: Weight Management, Peptide Therapy, Hormone Optimization, Longevity & Metabolic, Sexual Health, Immune & Inflammation, Hair Restoration, Dermatology.

Every tool accepts a medication as a name, id, or brand/alias (e.g. "Tirzepatide", "tirzepatide", "Mounjaro", or "copper peptide" → GHK-Cu). Unrecognized queries return "did you mean" suggestions.

Compounds covered

Semaglutide · Tirzepatide · Liraglutide · Naltrexone/Bupropion · BPC-157 · Sermorelin · CJC-1295/Ipamorelin · Ipamorelin · Tesamorelin · Thymosin Beta-4 (TB-500) · Testosterone (cypionate & cream) · Estradiol · Progesterone · DHEA · Anastrozole · Pregnenolone · hCG · NAD+ · Metformin · Rapamycin · Berberine · Resveratrol · NMN · PT-141 · Oxytocin · Tadalafil · Sildenafil · Thymosin Alpha-1 · Glutathione · Low-Dose Naltrexone · Finasteride · Oral Minoxidil · GHK-Cu · Tretinoin.

Install & build

git clone <repo> longevity-mcp-server
cd longevity-mcp-server
npm install

npm run build             # compile TypeScript → dist/ and copy data
npm run smoke             # end-to-end test (optional)

The clinical data ships as JSON in src/data/ and is copied into dist/data/ at build.

Use with Claude Desktop

Add to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "oak-longevity": {
      "command": "node",
      "args": ["/absolute/path/to/longevity-mcp-server/dist/index.js"],
      "env": { "LONGEVITY_LICENSE_KEY": "OAK-XXXX-XXXX-XXXX" }
    }
  }
}

Restart Claude Desktop. You can then ask things like:

• "List the peptide therapy options."
• "What's the dosing protocol for tirzepatide for weight loss?"
• "Can I prescribe tadalafil to a 60-year-old man taking nitroglycerin?"
• "Check interactions between rapamycin, simvastatin, and clarithromycin."
• "What baseline labs do I need before starting testosterone?"
• "A patient reports fatigue, low libido, and wants to lose weight — what pathways fit?"

See examples/claude_desktop_config.json for an npx variant.

Use with Claude Code

claude mcp add oak-longevity -- node /absolute/path/to/longevity-mcp-server/dist/index.js

Inspect locally

npm run inspect          # opens the MCP Inspector against the stdio server

Remote hosting (HTTP / SSE)

The same tools are served over Streamable HTTP for remote deployment (MCPize, a VPS, or serverless):

npm run build
PORT=3000 node dist/http.js
# → POST http://localhost:3000/mcp     (GET /health for a liveness check)

The HTTP transport is stateless and multi-tenant: the per-customer license key is read from a request header, so a single deployment can serve many customers.

X-Oak-License: OAK-XXXX-XXXX-XXXX        (preferred)
Authorization: Bearer OAK-XXXX-XXXX-XXXX (also accepted)

Monetization & licensing

The server has a built-in free / premium split designed to be wired to a billing provider (Stripe, MCPize) with minimal change.

• Free tier: get_medication_list, get_medication_details, get_fda_status — the catalog and regulatory reference.
• Premium tier: the clinical decision-support engine — dosing protocols, baseline labs, monitoring plans, contraindication screening, drug-interaction checks, and intake pathway suggestions.

Premium tools remain discoverable (they appear in tools/list so clients can advertise the upgrade), but calling one without a valid entitlement returns an upgrade prompt instead of data.

Entitlement resolution

Configured via environment variables (stdio) or request headers (HTTP):

A locally-issued key matches the format OAK-XXXX-XXXX-XXXX. For production, point LONGEVITY_LICENSE_VERIFY_URL at your billing webhook; it should accept { "key": "..." } and return { "valid": true, "tier": "premium", "expiresAt": "..." }.

The verification layer lives entirely in src/licensing.ts behind a LicenseProvider interface — swap the implementation without touching any tool.

Project structure

longevity-mcp-server/
├── src/
│   ├── index.ts          # stdio entry point (Claude Desktop / Code)
│   ├── http.ts           # Streamable HTTP entry point (remote hosting)
│   ├── server.ts         # builds the MCP server + tier gating
│   ├── licensing.ts      # free/premium entitlement (pluggable)
│   ├── data.ts           # data loading + medication resolver
│   ├── types.ts          # clinical data types
│   ├── tools/            # one file per MCP tool (9 tools)
│   └── data/             # clinical data (JSON)
│       ├── categories.json
│       ├── medications.json
│       ├── dosing.json
│       ├── contraindications.json
│       ├── interactions.json
│       ├── labs.json
│       ├── fda.json
│       └── pathways.json
├── scripts/
│   ├── copy-assets.mjs   # copy JSON into dist/ at build
│   └── smoke-test.mjs    # end-to-end MCP client/server test
├── examples/
│   └── claude_desktop_config.json
├── package.json
├── tsconfig.json
├── LICENSE
└── README.md

Data model

The clinical data is hand-curated from standard pharmacology references and longevity-medicine practice (Endocrine Society / Menopause Society / ISSWSH guidance, FDA labeling and shortage/bulk-substance lists, and the peer-reviewed literature for off-label and investigational compounds). Each dataset is keyed by medication id:

• medications.json — class, mechanism, formulations, candidate profile, schedule.
• dosing.json — per-indication route / start / titration / maintenance / max / evidence grade.
• contraindications.json — boxed warnings, absolute & relative contraindications (with machine-matchable condition keywords), cautions, pregnancy.
• interactions.json — per-drug interaction rules (severity, effect, management).
• labs.json — baseline panels and ongoing monitoring schedule.
• fda.json — approval status, schedule, 503A/503B compounding considerations, approved uses, references.
• pathways.json — 15 intake pathways mapping symptoms/goals → first-line + adjunct medications.

Because regulatory status (especially FDA drug-shortage listings and 503A bulk-substance eligibility for peptides) shifts frequently, treat get_fda_status output as a starting point and confirm against the current FDA database before compounding.

License

MIT © 2026 Keith Schmidt, MD — Oak Longevity Institute

The clinical reference content is provided for educational and decision-support purposes only and does not constitute medical advice. See the disclaimer in LICENSE.