OpenConstruction · MCP Integration

Introduction

What is MCP?

Model Context Protocol (MCP) standardizes how AI assistants connect to external tools and data sources. The OpenConstruction MCP server exposes read-only access to our Datasets, Models, Use Cases, Tools, and Guides catalogs through a small set of server tools.

Example

Ask: “Find 2024 datasets for safety monitoring.” → receive structured metadata with links back to the catalog entries.

MCP quick links

Capabilities

Available Tools

Core tool groups exposed by the OpenConstruction MCP server:

🔍 Search resources

search_models search_datasets search_use_cases

📊 Get details

get_model_details get_dataset_details

🛠️ Browse & guidance

list_tools get_best_practices

📈 Catalog statistics

get_catalog_stats

Setup

Install & Connect

Minimal steps to run the server locally and connect it to an MCP-compatible client (e.g., Claude Desktop).

Clone repository

git clone https://github.com/ruoxinx/openconstruction-mcp-server.git
cd openconstruction-mcp-server

Install dependencies
```
npm install
```
Build server
```
npm run build
```

Configure Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "openconstruction": {
      "command": "node",
      "args": ["/path/to/build/index.js"]
    }
  }
}

See: Claude Desktop MCP guide

Restart & verify
Reopen Claude Desktop and run.

Tip — Private GitHub data

If your catalogs are in a private repository, configure your server with a fine-grained GitHub PAT (Contents: Read) or a classic PAT with minimal repo read scope to fetch JSON safely. Never embed tokens in client config; store them in server-side environment variables.

Examples

Usage Scenarios

Dataset discovery: “List point cloud datasets for semantic segmentation.”

Model search: “Find models for safety monitoring published after 2023.”

Use-case exploration: “Show industry deployments of computer vision.”