Noisemodelling-mcp

NoiseModelling MCP Server

An MCP (Model Context Protocol) server that exposes NoiseModelling as AI-callable tools. Lets AI agents (Claude, etc.) run full urban noise mapping pipelines from raw OSM data to GeoJSON noise contour maps.

What it does

Wraps the NoiseModelling WPS API into ~25 structured MCP tools covering:

• Setup - download, install, and start NoiseModelling automatically
• Static noise maps - OSM import -> receiver grid -> CNOSSOS-EU propagation -> isosurface -> GeoJSON export
• Dynamic noise maps - per-timestep vehicle trajectories -> LA10/LA50/LA90 indicators -> QGIS temporal animation
• Database management - list, drop, and clean tables

Prerequisites

Requirement	Version
Python	>= 3.13
NoiseModelling	5.x or 6.x
Java	>= 11 (bundled with NoiseModelling 5.x)

NoiseModelling must be running before any analysis tool is called. Use the setup tools (below) to install and start it, or start it manually.

Installation

# Clone
git clone https://github.com/YuGaobjtu/Noisemodelling-mcp
cd Noisemodelling-mcp

# Install dependencies (uv recommended)
uv sync
# or: pip install -e .

Quick start

1 - Start the MCP server

uv run main.py
# Server listens on http://127.0.0.1:8000 (streamable-http transport)

2 - Configure your MCP client

Claude Desktop (claude_desktop_config.json):

{
  "mcpServers": {
    "noisemodelling": {
      "command": "uv",
      "args": ["run", "/path/to/Noisemodelling-mcp/main.py"]
    }
  }
}

Claude Code (.claude/settings.json):

{
  "mcpServers": {
    "noisemodelling": {
      "command": "uv",
      "args": ["run", "/path/to/Noisemodelling-mcp/main.py"]
    }
  }
}

3 - Install & start NoiseModelling (first time)

Ask your AI agent:

"Get the latest NoiseModelling release, download it, install it, and start the server."

Or call tools directly in order:

1. get_noisemodelling_release_info - get download URL
2. download_noisemodelling - save ZIP
3. install_noisemodelling - extract
4. start_noisemodelling_server - launch
5. check_server_status - verify

4 - Run a noise map

Run a full noise map pipeline for my_city.osm or demo data like:(NoiseModelling\NoiseModelling_5.0.1\data_dir\data\wpsdata\map.osm.pbf), SRID 2154,
output to /tmp/noise_output/noise_map.geojson

The agent calls run_full_noise_map_pipeline - one tool that runs all 6 steps.

Environment variables

Variable	Default	Description
NOISEMODELLING_URL	http://127.0.0.1:9580	Override NoiseModelling server URL

Copy .env.example to .env and edit as needed.

Tools reference

Setup tools

Tool	Description
get_noisemodelling_release_info	Fetch latest release info + download URLs from GitHub
download_noisemodelling	Download ZIP to disk
install_noisemodelling	Extract ZIP to install directory
start_noisemodelling_server	Launch server process in background
check_server_status	Verify server is reachable, list WPS processes

Static noise mapping

Tool	Output tables
import_osm	BUILDINGS, GROUND, ROADS
generate_receiver_grid	RECEIVERS, TRIANGLES
compute_road_emission	LW_ROADS
compute_noise_propagation	RECEIVERS_LEVEL
create_isosurface	CONTOURING_NOISE_MAP
export_table	file on disk
run_full_noise_map_pipeline	all of the above in one call

Dynamic noise mapping - Case 1 (single traffic flow)

Tool	Output tables
generate_regular_receiver_grid	RECEIVERS, TRIANGLES
flow_to_noisy_vehicles	SOURCES_GEOM, SOURCES_EMISSION
compute_dynamic_noise_propagation	RECEIVERS_LEVEL
compute_dynamic_indicators	RECEIVERS_LEVEL_DYN_IND (LA10/50/90)
run_dynamic_noise_pipeline_case1	all Case 1 steps in one call

Dynamic noise mapping - Case 2 (traffic at regular intervals)

Tool	Output tables
import_file	any table from GeoJSON/SHP/CSV
split_sources_period	SOURCES_GEOM, SOURCES_EMISSION
run_dynamic_noise_pipeline_case2	all Case 2 steps in one call

Database management

Tool	Description
list_database_tables	List all tables (optionally with columns)
export_all_tables	Export every table to a directory
drop_table	Remove one table
clean_database	Remove all tables

Platform support

Tested on Windows. Linux/macOS paths use ~/NoiseModelling as default root. NoiseModelling itself runs on all platforms with Java 11+.

License

MIT - see LICENSE.

Acknowledgements

Built on NoiseModelling by Universite Gustave Eiffel / UMRAE. Implements CNOSSOS-EU road noise model per EU END Directive 2002/49/EC.