ArcGIS MCP Server

ArcGIS MCP Server exposes ArcGIS Online search and feature-layer analysis through both HTTP and MCP.

Attribution

This repository was started from and is based on Garrick Garcia's ArcGISMCP.

Credit to Garrick Garcia for the original project foundation and upstream work.

Disclaimer

This project is "vibe coded": it was assembled rapidly with AI-assisted iteration and should be reviewed, tested, and validated before production use.

Setup

Prerequisites

• Python 3.11+
• uv
• An ArcGIS Online account

1. Install uv

If needed, install uv first:

macOS:

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows PowerShell:

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Create the Virtual Environment and Install Dependencies

uv sync

This creates .venv and installs the project plus its dependencies.

You do not need to activate the environment if you run commands through uv directly.

3. Configure Environment Variables

Create your local environment file from the example before you run the server.

macOS/Linux:

cp .env.example .env

Windows PowerShell:

Copy-Item .env.example .env

Windows Command Prompt:

copy .env.example .env

Then update .env with your ArcGIS credentials and any optional overrides:

ARCGIS_USERNAME=your_username
ARCGIS_PASSWORD=your_password
ARCGIS_URL=https://www.arcgis.com
ARCGISMCP_HOST=127.0.0.1
ARCGISMCP_PORT=8000
ARCGISMCP_CORS_ORIGINS=http://localhost:5173,http://127.0.0.1:5173

4. Run the Server

After .env is configured, start the server with uv:

uv run python main.py

When the server starts, these endpoints are available:

• REST API: http://127.0.0.1:8000/
• Swagger docs: http://127.0.0.1:8000/docs
• MCP endpoint: http://127.0.0.1:8000/mcp
• Health check: http://127.0.0.1:8000/health

/mcp is for MCP clients, not for manual browser testing.

What It Provides

• search_layers: find matching feature layers and return REST layer URLs
• get_feature_table: return a sample of attribute rows as CSV
• summarize_field: return stats and top values for one field
• search_content: search general ArcGIS content such as dashboards and web maps

Examples

Check the service

curl "http://127.0.0.1:8000/health"

Find feature layers

curl "http://127.0.0.1:8000/api/search/layers?keyword=Hydrants"

Example response shape:

{
   "keyword": "Hydrants",
   "count": 2,
   "matches": [
      {
         "name": "Fire Hydrants",
         "url": "https://services.arcgis.com/.../FeatureServer/0",
         "item_title": "Fire Hydrants"
      }
   ]
}

Preview a layer table

curl "http://127.0.0.1:8000/api/feature-table?service_url=https://services.arcgis.com/.../FeatureServer/0"

Summarize a field

curl "http://127.0.0.1:8000/api/field-summary?service_url=https://services.arcgis.com/.../FeatureServer/0&field_name=Status"

Search general ArcGIS content

curl "http://127.0.0.1:8000/api/search/content?keyword=Traffic&item_type=Dashboard"

Common Flow

1. Call search_layers or /api/search/layers with a keyword.
2. Copy a returned feature layer URL.
3. Call get_feature_table or /api/feature-table to inspect fields.
4. Call summarize_field or /api/field-summary for a field you care about.

Notes

• The server authenticates lazily using the credentials in .env.
• get_feature_table returns a sample, not the full dataset.
• Local frontend CORS is enabled by default for Vite-style localhost origins.