Mediora.AI MCP

Production Model Context Protocol server exposing Mediora's doctor-reviewed medical knowledge graph as agent-callable tools — blood-test markers, clinical conditions, patient symptoms, and Israeli sick funds.

Protocol 2025-03-26 Server v0.5.0 Streamable HTTP EN · RU · HE No PHI in read surface JWT for 4 write tools 14 tools
https://mcp.mediora.ai/mcp

Add to Claude Desktop

Append this block to claude_desktop_config.json (macOS ~/Library/Application Support/Claude/, Windows %APPDATA%\Claude\):

{
  "mcpServers": {
    "mediora": {
      "type": "http",
      "url": "https://mcp.mediora.ai/mcp"
    }
  }
}

Quick probe with curl

curl -X POST -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
  https://mcp.mediora.ai/mcp

What the server exposes

14 tools

NamePurpose
mediora__list_markersList every blood-test marker in the Mediora catalog.
mediora__explain_markerReturn the full doctor-reviewed explainer for a single blood-test marker by slug.
mediora__list_conditionsList every clinical condition in the Mediora catalog.
mediora__explain_conditionReturn the full doctor-reviewed explainer for a clinical condition by slug.
mediora__lookup_markerFuzzy-search markers by alias across EN/RU/HE.
mediora__list_symptomsList every patient-side symptom in the Mediora catalog (fatigue, joint pain, headache, etc.
mediora__explain_symptomReturn the full doctor-reviewed triage explainer for a single symptom by slug.
mediora__list_sick_fundsList the four Israeli sick funds (Clalit, Maccabi, Meuhedet, Leumit) — the HMOs that distribute every Israeli citizen's primary medical coverage.
mediora__explain_sick_fundReturn the full explainer for a single Israeli sick fund by slug.
mediora__whoamiValidate a Mediora.
mediora__analyze_lab_pdfUpload a lab-report PDF or scan (by HTTPS URL) and trigger Mediora.
mediora__get_patient_historyReturn the patient's past medical-test rows (id, test type, test date, risk level, summary, status, abnormal-marker count).
mediora__get_test_detailsReturn the persisted markers + AI analysis (Summary, KeyFindings, Recommendations, DetailedExplanation) for a specific medical-test the authenticated patient owns.
mediora__longitudinal_trendReturn the per-marker trajectory analyser output for the authenticated patient — HbA1c rising, eGFR falling, ferritin depleting, etc.

What's inside the catalogs

CatalogEntriesLanguages
Blood-test markers65 (ferritin, HbA1c, TSH, ANA, vitamin D, ...)ENRUHE
Clinical conditions32 (type-2 diabetes, hypothyroidism, Addison's, ...)ENRUHE
Patient symptoms26 (fatigue, palpitations, blurred-vision, ...)ENRUHE
Israeli sick funds4 (Clalit, Maccabi, Meuhedet, Leumit)ENRUHE

Authentication model

Read tools (catalog lookup, marker / condition / symptom / sick-fund queries) are anonymous — no token, no API key. You pass the JSON-RPC request and you get the data.

Four write tools — mediora__whoami, mediora__analyze_lab_pdf, mediora__get_patient_history, mediora__longitudinal_trend — are gated by a patient JWT. That's the same token a logged-in user has in localStorage.auth_token after signing in at mediora.ai. Pass it as arguments.bearer_token on the JSON-RPC call. The server validates against the same signing key the patient API trusts.

There is no server-level API key: anonymous clients see the public catalog; authenticated clients can additionally see their own data. We never collect a key to use the read surface.

Same data via the website

Every catalog entry that this MCP server returns is also rendered as a public HTML landing page at mediora.ai — e.g. /en/markers/ferritin (also /ru/ and /he/). MCP and HTML share the exact same catalog JSON under Mediora.Core/Domain/.

License + Citation

Catalog content (markers, conditions, symptoms, sick funds) is CC-BY-NC 4.0 — free to share with attribution and a link back, non-commercial unless licensed. Code is MIT once the public mirror lands at github.com/SLAtech-IL/mediora-mcp. Cite as Mediora.AI with link to mediora.ai.

Discovery manifest

Machine-readable server manifest is at /.well-known/mcp.json; quick GET envelope (this server's metadata as JSON) is at /mcp with method GET. Send JSON-RPC over POST to the same /mcp URL.

Scope note

Mediora.AI is a medical lab-report interpretation SaaS targeting the IL / RU / EU markets. This MCP server is the public-data-only slice of that system; it does not expose PHI, patient identifiers, or proprietary analysis logic. The website at mediora.ai runs the authenticated patient pipeline (PDF upload → OCR → AI interpretation → encrypted at rest); only the catalog reference data is mirrored here.