# Zhonghao1995/agentic-swmm-workflow [Health: Active]

**Category:** 🌳 Environment & Nature  
**Repository:** https://github.com/Zhonghao1995/agentic-swmm-workflow  
**GitHub Stars:** 27  
**npm Downloads (last month):** 38365913  
**Views:** 3  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/zhonghao1995-agentic-swmm-workflow

## Description
Eleven MCP servers exposing a reproducible EPA SWMM stormwater-modelling workflow: model building, simulation runs with manifests and continuity checks, calibration, GIS/QGIS integration, design storms and climate scenarios, uncertainty analysis, plotting, and modelling memory. Config generators included for Codex, Claude Code, OpenClaw, and Hermes.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "agentic-swmm-workflow": {
    "command": "npx",
    "args": ["-y","skills"]
  }
}
```

## Documentation

## What Zhonghao1995/agentic-swmm-workflow MCP server does

Zhonghao1995/agentic-swmm-workflow MCP server provides an agentic layer around EPA SWMM stormwater modelling. It covers the stages commonly needed for a project: GIS preprocessing, rainfall preparation, model construction, input-file generation, simulation, quality assurance, plotting, calibration, uncertainty analysis, reporting, and provenance tracking.

The repository describes eleven MCP servers organized around these workflow areas. The broader aiswmm runtime can accept modelling goals in natural language, but the resulting work remains grounded in files and command-line actions. Outputs can include SWMM input, report, and binary files; manifests; command traces; QA summaries; parsed peak-flow metrics; figures; calibration summaries; uncertainty summaries; and audit records such as `experiment_provenance.json`.

## How it works

The workflow separates execution, modelling memory, and controlled skill evolution. Execution handles reproducible SWMM actions and produces artifacts that can be inspected or reused. Modelling memory records recurring issues and workflow knowledge. Repeated patterns may lead to proposals for improving skills, but proposed changes require human review and benchmark verification.

Simulation runs are designed to include manifests and continuity checks, so model results can be reviewed alongside the conditions and commands that produced them. The project also supports deterministic SWMM execution and audit stages before outputs are treated as evidence. GIS and QGIS-related processing, design storms, climate scenarios, calibration, and uncertainty work are exposed as separate workflow capabilities rather than being hidden inside one conversational operation.

Zhonghao1995/agentic-swmm-workflow MCP server can also be used with agent runtimes other than its built-in aiswmm runtime. The README specifically names Codex, Claude, OpenClaw, and Hermes, and provides configuration generators for those environments.

## Setup and configuration

The project is open source and licensed under MIT. The README identifies version 0.9.4 as a stable pre-1.0 release and documents installation through the `aiswmm` Python package, one-line macOS/Linux and Windows installers, and Docker. It also provides a pinned Docker example using `ghcr.io/zhonghao1995/agentic-swmm-workflow:v0.9.4` and mounts a local `runs` directory for generated run data.

After installation, the runtime is launched with `aiswmm`. An `aiswmm setup` wizard presents provider routes and detects available services. The material states that some routes do not require an API key, while other provider choices may involve credentials. API keys should be configured through the documented configuration process rather than pasted into the runtime conversation.

## Tools and capabilities

Zhonghao1995/agentic-swmm-workflow MCP server supports:

- Building and preparing SWMM models
- Running simulations with manifests and continuity checks
- Auditing runs and recording provenance
- Connecting GIS and QGIS preprocessing to modelling workflows
- Creating design-storm and climate-scenario workflows
- Calibrating models and summarizing uncertainty
- Producing rainfall-runoff plots and other modelling figures
- Maintaining modelling memory and proposing supervised skill refinements

## Limitations and notes

The project is pre-1.0, so interfaces and behavior may change. It is not intended to replace EPA SWMM or the modeller. Natural-language orchestration does not remove the need to inspect model files, simulation outputs, QA results, and audit artifacts. Skill-evolution proposals are not applied automatically: human review and benchmark verification remain required.

The README describes a built-in runtime and multiple MCP interfaces, but the supplied material does not specify a single standalone command for launching each MCP server. It also does not establish compatibility with every MCP client or identify a required environment-variable set.

_Full upstream README: https://allmcps.com/mcp/zhonghao1995-agentic-swmm-workflow/readme_