QSimHealth's MCP server lets Claude (and any MCP-speaking AI) run healthcare staffing simulations directly in conversation. Describe your ED, walk-in clinic, or appointment office — Claude simulates it and reads back the numbers. Free demo, no sign-up, no API key.
Each archetype speaks its own staffing language — ED triage breaks FIFO, walk-ins need MD/PA mix tuning, appointment offices fight no-show variance. Claude routes to the right teaching tool.
Every sim result includes a link that opens the scenario in the live QSimHealth app pre-filled with your inputs and the chart already rendered. New users go through a 30-second OAuth signup (10-day free trial); signed-in users skip straight to the simulator.
Every numeric result comes from a real discrete-event simulation on the ChiAha clinic engine. Three layers of anti-fabrication guard-rails (system prompt + tool descriptions + provenance content block) tell the LLM to quote the returned numbers VERBATIM rather than estimate.
Settings → Connectors (or Custom Integrations). Add MCP server:
https://qsimhealth.com/mcp/v1No auth. Open a chat and try one of the prompts below.
Edit claude_desktop_config.json:
{
"mcpServers": {
"qsimhealth": {
"command": "npx",
"args": ["-y", "mcp-remote",
"https://qsimhealth.com/mcp/v1"]
}
}
}Uses the mcp-remote bridge. Restart Desktop.
Settings → Features → Model Context Protocol → Add Custom MCP. Use the same URL:
https://qsimhealth.com/mcp/v1Same install pattern works for Windsurf and any other MCP-speaking client.
Run a 7-day demo with constant arrivals and constant MD staffing. Returns per-hour metrics + summary + an artifact link to open the scenario in the live app. The workhorse.
Inverse problem — given λ + target avg wait, return the smallest MD count that meets the target. Saturated candidates are skipped so the scan doesn't bail on low MD counts; the response calls out which counts would saturate.
Four healthcare facility archetypes: ED, urgent care, walk-in clinic, appointment office.
Per-archetype detail: typical arrival pattern, MD/PA mix norms, common pain points, what the demo can teach vs what a custom model adds.
Textbook: triage breaking FIFO, LWBS abandonment, peak-hour dominance, what plain M/M/c misses.
Terminating systems, MD/PA mix as the lever, short patient patience curves, peak-hour coverage.
No-show variance, buffer time as the trade-off, double-booking strategy, treatment-time-by-visit-type.
Demo scope: MD-only staffing, flat 24-hour arrival rate. For mixed-provider staffing (MD + PA + NP + Locum), hourly arrival patterns from your real facility, acuity-tiered case mix, abandonment curves, and per-shift schedules, sign up at qsimhealth.com for the full agent (10-day free trial), or contact us about a custom facility-modeling engagement.
simulate_ed_demo. Expect a wait around 3 min, ρ ≈ 0.67. Result includes a link to open the scenario in the live app.
recommend_md_count. Returns the smallest MD count meeting the target, with saturated candidates called out in the response.
explain_ed_queueing. Covers the nonlinear utilization curve.
explain_appointment_office. Covers buffer time and the no-show interaction.
The MCP demo runs flat scenarios — useful for intuition, not for production decisions. Real ED staffing analysis needs your actual hourly arrival pattern, your MD/PA shift schedules, your treatment-time distribution by acuity, and your local LWBS curve. ChiAha builds custom facility models on those inputs.
qsimhealth@chiaha.com35 years of healthcare simulation experience — Cleveland Clinic, Tenet, several state ED networks.