A
licenseA
qualityD
maintenanceEnables AI assistants to interact with Autodesk Civil 3D, allowing them to retrieve project data, create/modify/delete drawing elements, and execute code to automate Civil 3D operations.
3
44
MIT
Garden Design MCP Server
Allows using OpenStreetMap as a map background under the garden, helping locate and visualize the plot against real-world maps.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Garden Design MCP ServerHelp me design my garden step by step."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Garden Design MCP Server
Let your AI assistant design a garden on your real plot. This MCP (Model Context Protocol) server connects Claude Code, Claude Desktop, Cursor or any other MCP client to the Zaprojektuj-Ogrod.pl Garden Editor, a browser app for garden and landscape design in 2D and 3D. You watch every change appear in the editor as the assistant makes it, then walk through the finished garden in 3D.
Step-by-step setup guide: in English · po polsku
What the assistant can do
Find the plot anywhere in the world, on OpenStreetMap or a satellite photo. In Poland it imports the exact parcel boundary and the buildings from the land registry (Geoportal, ULDK).
Read your photos and floor plan to put the doors, windows and views on the right walls of the house.
Draw what is already there: fences, paving, the driveway, a heat pump on the wall. Nothing that exists is moved without asking.
Lay out zones for your wishes at practical sizes (a terrace for six, a path, raised beds, a bench corner), and say so when a wish does not fit instead of shrinking it quietly.
Choose plants from a catalog of 2,000+ by sun, soil, hardiness and size, and place them at their mature spread. Surfaces, furniture, raised beds and planters come from the same catalog.
Check its own work: planting density per bed, previews of the plan, 2D and 3D views. It saves the design to your account.
It works one step at a time and stops for your sign-off: the site first, then the zones, then the planting.
Related MCP server: Civil 3D MCP Server
Quick start
1. Add the server to your AI client. No install is needed: npx downloads and runs it.
Claude Code (~/.claude.json):
{
"mcpServers": {
"garden-editor": {
"command": "npx",
"args": ["@zaprojektujogrod/mcp-server"]
}
}
}Claude Desktop (Settings → Developer → MCP Servers):
{
"garden-editor": {
"command": "npx",
"args": ["@zaprojektujogrod/mcp-server"]
}
}2. Open a garden in the Garden Editor at zaprojektuj-ogrod.pl and switch the MCP connection on in Settings › AI assistant. The editor connects to the server on your computer.
3. Ask your assistant, for example: "Help me design my garden, step by step." Or start from the server's start_design prompt.
How it works
┌─────────────────┐ stdio ┌──────────────────┐ WebSocket ┌──────────────────┐
│ AI client │ ◄────────────► │ MCP server │ ◄──────────────► │ Garden Editor │
│ (Claude, etc.) │ │ (this package) │ localhost:3333 │ (your browser) │
└─────────────────┘ └──────────────────┘ └──────────────────┘The server runs on your computer and talks to the editor open in your browser over a local WebSocket (ws://localhost:3333). Your garden stays in your account; the assistant runs in your own AI client, on your own subscription or API key.
Setup from source
If you prefer to run from source instead of npx:
git clone https://github.com/zaprojektuj-ogrod/garden-design-mcp.git
cd garden-design-mcp
npm install
npm run buildThen configure your AI client with a direct path:
{
"mcpServers": {
"garden-editor": {
"command": "node",
"args": ["/path/to/garden-design-mcp/dist/index.js"]
}
}
}Available tools (63)
Tools follow the design funnel: Site → Context → Zones → Detailing → Review. Most tools that create or change things have a batch sibling (add_objects, update_objects, delete_objects, add_pins, add_zones); prefer the batch form when placing many items. Call get_design_guide once at the start: it holds the whole workflow, the photo method and the planting rules.
Phase 1: Site
Tool | Description |
| Find a place anywhere in the world from an address |
| Place the garden on the map at real-world coordinates |
| Get the garden's place |
| Show or hide a map under the garden: OpenStreetMap, the satellite photo (Poland) or a laid survey map |
| Give the plot boundary as a polygon in metres (from measurements or traced from the map) |
| Land registry parcel by WGS84 lat/lon (Poland) |
| Land registry parcel by parcel ID (Poland) |
| Address → land registry parcel (Poland) |
| Import a parcel as the garden boundary, with the cadastral map behind it (Poland) |
| Import the buildings on and around the parcel from the land registry (Poland) |
| House facades by compass side, party walls and the garden depth on each side |
| Upload a reference photo (kept for the session) |
| Get the uploaded reference photos |
| A close-up of an image attached in the editor's chat, for reading details |
| Lay a survey map or site plan to scale under the garden, from two matched points or two points and a distance |
| Turn points on the laid survey map into garden coordinates in metres |
Phase 2: Context
Tool | Description |
| Add one or many context pins (door, window, gate, view) |
| Delete a pin by ID |
| Get all pins |
| Get garden metadata (name, rotation, boundary, stats) |
| Set north orientation in degrees |
| Set the garden's free-text description (the assistant's memory between sessions) |
| Capture a top-down view as an image the assistant sees inline |
Phase 3: Zones
Tool | Description |
| Add one or many zones (polygon or path in metres); each answers with area, narrowest width and overlaps |
| Update zone properties |
| Delete a zone by ID |
| Get all zones |
| Get available zone/pin types with colors and labels |
| List objects whose centers fall inside a zone |
| Density and coverage of a zone (planted area %, objects by category) |
| Clear scene contents: |
| Export zones and pins as JSON |
| Import zones and pins from JSON |
Phase 4: Detailing
Tool | Description |
| Search the product catalog by layer and name and/or |
| Filter plants by sun, soil, hardiness, flower color, height, spread, bloom time, plus an optional name query |
| The valid filter values for |
| A plant's full description: characteristics, requirements, maintenance |
| Top-level categories for a layer |
| Leaf categories under a parent category |
| Full product details by ID |
| Get all garden objects (filter by layer; |
| Add one or many plants, architecture or ground surfaces ( |
| Update one or many objects (position, size, rotation, product) |
| Delete one or many objects by ID |
Phase 5: Review
Tool | Description |
| Snapshot of the scene; |
| Switch between 2D and 3D view |
| Get current view mode |
| Show/hide a layer |
| Get visibility state of all layers |
| Focus the camera on an area |
| Save all changes to the account |
| Undo last action |
| Redo last undone action |
| Coordinate system docs (units, axes, scale) |
Utility
Tool | Description |
| The full design guide: workflow, photo method, zone sizes, journal, plant selection |
| Check if the browser is connected |
| Set coordinate units to |
|
|
capture_preview: what the assistant sees
capture_preview returns an MCP ImageContent block ({type: "image", data, mimeType: "image/jpeg"}). The connected assistant sees it directly and can describe and check the plan without any file on disk.
It does not write a file to disk. MCP clients (Claude Desktop, Claude Code, Cursor, …) show the image inline, and the server does not keep screenshots. To keep the JPEG, right-click → "Save image as…" on the image in your client, or read the base64 in content[0].data from your own client harness.
MCP resources
URI | Description |
| Garden metadata |
| All zones as JSON |
| All context pins |
| All garden objects |
| Coordinate system documentation |
MCP prompts
Prompt | Description |
| Begin a new garden design, as the editor's guided set-up does: the site first, then zones and planting, each signed off |
| Resume an existing garden (assess state, suggest next steps) |
| Review and optimize the zone layout within the boundary |
Development
Run in development mode (with hot reload):
npm run devThe server logs to stderr so it does not interfere with MCP stdio communication.
Po polsku
Serwer MCP do projektowania ogrodów z asystentem AI. Łączy Claude, Cursor lub inny klient MCP z edytorem Zaprojektuj-Ogrod.pl. Asystent znajduje działkę (w Polsce z granicami i budynkami z Geoportalu), rysuje to, co już na niej jest, układa strefy według Twoich życzeń i dobiera rośliny z katalogu, a Ty widzisz projekt ogrodu w 2D i 3D. Instrukcja: Konfiguracja serwera MCP.
License
MIT, see LICENSE. Questions and bug reports: zaprojektuj-ogrod.pl/kontakt.
This server cannot be deployed
Maintenance
ActivityMaintained
ResponsivenessNo issues
Related MCP Connectors
Identify plants, mushrooms and insects, and keep your garden’s records and jobs, from your AI.
Design a house, its rooms and its garden in AI Plan Creator by asking in a chat.
AI interior, exterior and landscape design, photoreal rendering, image editing, video and 3D.
Design homes with your AI: floor plans, furniture, joinery, lighting and photos, live.
Related MCP Servers
- AlicenseAqualityCmaintenanceEnables AI assistants to interact with Autodesk Civil 3D through natural language, supporting tools for surfaces, alignments, profiles, corridors, pipe networks, COGO points, and AutoCAD geometry.93MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants to query SketchUp models, extrude massing volumes, configure solar shadows, and capture viewport renderings through MCP tools.MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI assistants to directly operate ArcGIS Pro through natural language, giving access to 2,500+ spatial analysis tools across 46 toolboxes via the Model Context Protocol.MIT