01 / 06
Vital sign arithmetic
vital_signs_summaryPulse pressure, estimated MAP, and shock index—returned as calculations, never a risk judgment.
MAP ≈ DBP + ⅓(SBP − DBP)
OPEN-SOURCE · LOCAL-FIRST · MCP
Clinical calculations.
Callable by AI.
Six deterministic calculation and conversion tools for AI clients that support local MCP servers. Clear inputs. Explicit units. No clinical interpretation.
Local stdio transport Python 3.11+MIT licensed
ONE INPUT. ONE DETERMINISTIC RESULT.
mL / hr125.00clear units
TRANSPORTstdioruns locally
$
six tools · GitHub mainuvx --from git+https://github.com/Umarjaum/clinical-calc-mcp.git clinical-calc-mcpTHE TOOLKIT
Small tools.
Clear answers.
Purpose-built functions that return the requested arithmetic—without silently crossing into diagnosis, triage, or treatment advice.
02 / 06
Parkland formula
parkland_formulaReproducible 24-hour fluid arithmetic with first-eight-hour and remaining-volume breakdowns.
4 mL × kg × %TBSA
03 / 06
BSA & BMI
bsa_mostellerMosteller body-surface-area arithmetic and BMI output, without category or diagnosis.
√((cm × kg) / 3600)
04 / 06
Infusion rate
drip_rate_calculatormL/hour and drops/minute with explicit units and documented half-up whole-drop rounding.
mL ÷ hours · gtt/min
05 / 06
Temperature
temperature_converterCelsius ↔ Fahrenheit conversion with absolute-zero validation.
°C ↔ °F
06 / 06
Weight
weight_converterKilogram ↔ pound conversion with positive-value and finite-number checks.
kg ↔ lb
Finite-number, range, physical-boundary, and unit validation are part of every calculation.
A SIMPLE BOUNDARY
Your client.
Your machine.
The MCP server runs as a local process. Your AI client calls a named tool with structured inputs; the server returns deterministic values over stdio.
How MCP works01AI clientClaude · Cursor · VS Code
JSON-RPC · stdio
02Local MCP serverValidated tool inputs
deterministic output
03CalculationNumber + unit · no interpretation
No calculation-time network calls. Client privacy policies still apply.
MEET YOUR CLIENT WHERE IT IS
One server.
Compatible clients.
Use any MCP client that can launch a local stdio process. Setup menus and configuration-file locations vary by application and version.
CHOOSE YOUR TRACK
Published package or latest source?
The newest six-tool source is available on GitHub. Use uvx for a clean, isolated install without adding the package globally.
uvx --from git+https://github.com/Umarjaum/clinical-calc-mcp.git clinical-calc-mcpBoth choices run a local stdio MCP server. The PyPI package will switch to six tools after 0.2.0 is released. Read the complete quick start and troubleshooting guide
{
"mcpServers": {
"clinical-calc-mcp": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/Umarjaum/clinical-calc-mcp.git",
"clinical-calc-mcp"
]
}
}
}THE PROJECT, OPENLY DOCUMENTED
Find your next step.
Source, setup, safety, packaging, releases, and community links—all in one place.
SETUP GUIDE
Quick start & client setup
Choose an install track, connect Claude Desktop, Cursor, VS Code, and troubleshoot.
Open resourceREFERENCE
All six tool specifications
Inputs, units, outputs, formula notes, ranges, and validation details.
Open resourceSAFETY
Clinical safety notes
Intended use, limitations, calculation sources, and independent verification.
Open resourceGITHUB
Source & release history
Browse source, tests, tagged releases, and the current package changelog.
Open resourcePYTHON PACKAGE
PyPI package
See the current published version and distribution files.
Open resourceJOIN IN
Community & contribution
Ask questions, suggest improvements, or help improve documentation.
Open resourceSEO & SEARCH
Search visibility & site verification
Verify ownership with Google and Bing, submit the sitemap, and measure discoverability.
Open resourceHeads up This is a local stdio server, not a hosted HTTP endpoint. Remote-only AI platforms need a separately hosted, secured service to connect. We do not claim universal AI compatibility.
BUILT WITH A CLEAR BOUNDARY
Arithmetic is not
clinical judgment.
This software calculates user-supplied values. It is not validated clinical decision support and does not diagnose, triage, prescribe, or determine whether a result is appropriate for an individual. Always independently verify calculations and follow current local protocols.
No diagnosis or treatment advice Units and validation made explicit No patient data stored by the server
GOOD TOGETHER
Built in the open.
Better with you.
Whether you write Python, improve docs, test an MCP client, or spot an edge case, there’s room to help. Contributors are welcome from across the web.
Use synthetic examples only. Never post patient data, secrets, or confidential information.
GOOD TO KNOW
Frequently
asked.
Can I use it with any AI platform?
Only AI clients that support MCP and launching a local stdio process can use this version directly. A browser-only or remote-only platform would need a separately hosted and secured MCP endpoint.
Does it give medical advice?
No. It performs only the documented calculations from supplied numbers. It does not diagnose, triage, recommend treatment or dosing, or evaluate clinical suitability. Follow local protocols and independently verify results.
Which version is on PyPI?
PyPI currently publishes 0.1.0 with three tools. The six-tool 0.2.0 source is available on GitHub main, but is not yet published to PyPI. This page’s install selector shows which track you are choosing.
How do I contribute safely?
Read the contribution guide and use the GitHub issue forms. Use made-up test values only—never include patient identifiers, real clinical narratives, credentials, or other confidential data.
START WITH A FUNCTION