Skip to content

Latest commit

 

History

5 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

RWS: Rapid HVAC Workflow System

Custom air-handling-unit design and estimation as seven cooperating agents over four calculation servers. The engineering method is written in plain language; the arithmetic is done by code. Built on Claude Code and the Model Context Protocol, December 2025.

Status Specification, four MCP servers and a scripted demonstration, December 2025.
Agents 7 Claude Code skills: conductor, psychrometrics, airflow, thermal, design, cost, QA
Servers 4 MCP servers, TypeScript, 1,481 lines: psychrometrics, component-db, simulation, estimation
Contracts 3 JSON schemas: request, constraint, result
Demonstration examples/demo-workflow.md: a hospital surgery suite (100 % outdoor air, HEPA, NC 35, Houston climate). The script is here; no run is recorded
Licence MIT

A standalone project on this account: the agent-and-tool pattern from the governance architecture, applied to a regulated engineering discipline.

The problem

A custom air handler needs psychrometrics, coil and fan selection, acoustics, code compliance and cost estimation, and that knowledge sits with senior engineers. Encoding it as documents an agent applies, with the calculations done by deterministic servers, makes it reviewable, repeatable and transferable.

How it works

 request (CFM, supply temperature, climate, filtration, coil type)
     │
     ▼
 ahu-conductor ── validates the request, derives constraints, sequences the agents
     ├── ahu-design    conceptual design: unit arrangement and casing
     ├── ahu-psychro   mixed-air and coil-leaving states            → psychrometrics
     ├── ahu-thermal   coil selection, rows, water-side hydraulics  → component-db, simulation
     ├── ahu-airflow   face velocity, static pressure, fan selection → component-db, simulation
     ├── ahu-cost      bill of materials and price roll-up           → estimation
     └── ahu-qa        checks the result against the schema and the constraints
     │
     ▼
 result: a complete design with BOM and pricing, and the reasoning at each step
Server Tools Purpose
psychrometrics calculate, mix, process air properties, mixing, cooling and heating processes
component-db lookup, list, coils, fans fan curves and coil performance data; designed to connect to a real catalogue
simulation size, airflow, thermal sizing, airflow and thermal calculations, pressure drop
estimation price, component_price cost roll-up and bill of materials

Every design passes schema validation before it is returned: supply airflow within range, every required component present, a complete BOM. A skill is a text file; changing the selection method is an edit, not a retraining.

Running the demonstration

Requires Claude Code and Node.js 18+.

git clone https://github.com/toneron2/RWS.git && cd RWS
npm install
claude
> Design an AHU: 10,000 CFM, 55°F supply, Houston TX, chilled water cooling, MERV 13 filters

QUICKSTART.md has the step-by-step setup; docs/BUSINESS_CASE.md the cost case for an engineering team.

Extending it

Add an equipment type by writing a skill for its selection method, adding its data to component-db, and extending the schemas. The component-db server is the intended seam to an ERP, a vendor database or a pricing sheet.

Contact

Tony Slosar · TODOMODO.IO AGENCY LLC · anthonyslosar@gmail.com · t.me/toneron2 · slosars.me

About

Air-handling-unit design and estimation as seven cooperating agents over four calculation servers. Claude Code and MCP, December 2025.

Topics

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages