Skip to main content
Glama
shughestr
by shughestr
README.md
# Vancam MCP Server

<!-- mcp-name: io.github.shughestr/vancam-mcp -->

[Vancam.ai](https://vancam.ai) — **Traffic Cameras & Road Conditions** — as an [MCP (Model Context Protocol)](https://modelcontextprotocol.io) server. Gives AI agents live access to the same camera network behind Vancam's road-condition data: over 1 million traffic cameras worldwide, searchable by map bounds, radius, route, or nearest point.

- šŸ” **Search cameras** by bounding box, radius, route corridor, or nearest-to-point
- šŸ“ø **Fetch live frames** by camera asset ID, returned directly in the tool result
- 🌐 **Real-time data** from the same backend that powers the [Vancam.ai](https://vancam.ai) map
- šŸ¤– **AI-ready** — built with [FastMCP](https://github.com/modelcontextprotocol/python-sdk), works with Claude Desktop, Claude Code, and any MCP-compatible client

## Quick Start

**Option A: Install from PyPI**

```bash
uvx vancam-mcp   # or: pip install vancam-mcp
```

**Option B: Clone and install dependencies**

```bash
git clone https://github.com/shughestr/vancam-mcp.git
cd vancam-mcp
pip install -r requirements.txt
```

**2. (Optional) Set up an API key**

Requests work out of the box using a shared, rate-limited key (1 req/s, 500/month, pooled across all anonymous users). For higher limits, grab a free personal key from your [Vancam.ai account page](https://vancam.ai) and set it as an environment variable:

```bash
cp .env.example .env   # then edit .env
```

```bash
# .env
VANCAM_API_KEY=your_personal_key_here
```

**3. Register the server with your MCP client**

For Claude Desktop or Claude Code, add to your MCP config. If installed from PyPI (Option A):

```json
{
  "mcpServers": {
    "vancam": {
      "command": "uvx",
      "args": ["vancam-mcp"],
      "env": {
        "VANCAM_API_KEY": ""
      }
    }
  }
}
```

If running from a local clone (Option B, see `.mcp.json` in this repo):

```json
{
  "mcpServers": {
    "vancam": {
      "command": "python3",
      "args": ["/absolute/path/to/vancam-mcp/vancam_mcp/server.py"],
      "env": {
        "VANCAM_API_KEY": ""
      }
    }
  }
}
```

`VANCAM_API_KEY` is optional — leave it blank to use the shared, rate-limited key.

Restart your client, and the tools below become available.

## Available Tools

### `list_cameras`
List cameras within a bounding box — the same query the VanCam map runs on pan/zoom.

```python
list_cameras(min_lat=49.2, min_lon=-123.2, max_lat=49.3, max_lon=-123.0, limit=50)
```
| Parameter | Type | Description |
|---|---|---|
| `min_lat`, `min_lon`, `max_lat`, `max_lon` | float | Bounding box (WGS84) |
| `limit` | int, optional | Max results, 1–100 (default 100) |
| `active_only` | bool, optional | Only `camera_class=open` live feeds (default false) |

### `get_cameras_by_radius`
Get cameras within a radius of a point.

```python
get_cameras_by_radius(lat=49.28, lon=-123.12, radius=1.0, limit=20)
```
| Parameter | Type | Description |
|---|---|---|
| `lat`, `lon` | float | Center point (WGS84) |
| `radius` | float, optional | Radius in km (default 1.0) |
| `limit` | int, optional | Max results (default 50) |
| `active_only` | bool, optional | Only open/live cameras |

### `get_cameras_along_route`
Get cameras along a straight-line corridor between two points.

```python
get_cameras_along_route(
    origin_lat=49.2827, origin_lon=-123.1207,
    dest_lat=49.1666, dest_lon=-123.1367,
    buffer=200.0, limit=50
)
```
| Parameter | Type | Description |
|---|---|---|
| `origin_lat`, `origin_lon`, `dest_lat`, `dest_lon` | float | Route endpoints |
| `buffer` | float, optional | Corridor width in meters (default 100.0) |
| `limit` | int, optional | Max results (default 50) |
| `active_only` | bool, optional | Only open/live cameras |

Results are sorted by `route_fraction` (0 = origin, 1 = destination). Note: this is a straight line between the two points, not a driving route.

### `get_nearest_cameras`
Get the closest cameras to a point.

```python
get_nearest_cameras(lat=49.2827, lon=-123.1207, limit=5)
```
| Parameter | Type | Description |
|---|---|---|
| `lat`, `lon` | float | Query point (WGS84) |
| `limit` | int, optional | Number of cameras (default 5) |
| `active_only` | bool, optional | Only open/live cameras |

### `get_camera_image`
Fetch a camera's live frame by asset ID, returned as image data in the tool result (not just a URL — the image endpoint requires an API key header that most MCP clients can't attach themselves).

```python
get_camera_image(asset_id="30145")
```

### `describe_camera_api`
Returns documentation for all search modes, camera fields, and image URL patterns. Call this first if you're unsure which tool to use.

## API Reference

Every search tool queries the same spatial API that backs the [Vancam.ai](https://vancam.ai) map:

| Purpose | URL |
|---|---|
| Spatial search | `https://api.vancam.ai/cameras/cameras` |
| Live image | `https://api.vancam.ai/api?asset_id={id}` |

Each camera includes `asset_id`, `latitude`, `longitude`, `street_address`, `direction`, `camera_class` (`open`/`premium`), `level1`/`level2`/`level3` (country/state/city), `distance_meters` (radius/nearest searches), `route_fraction` (route search), and `image_url`/`image_urls`.

Full schema: [`openapi.yaml`](openapi.yaml).

**Environment overrides:** `VANCAM_API_KEY`, `VANCAM_CAMERAS_SEARCH_URL`, `VANCAM_API_IMAGE_URL`

## Project Structure

```
vancam-mcp/
ā”œā”€ā”€ vancam_mcp/
│   ā”œā”€ā”€ server.py      # MCP server — registers the tools above
│   └── camera_api.py  # api.vancam.ai client
ā”œā”€ā”€ openapi.yaml       # API specification
ā”œā”€ā”€ pyproject.toml     # Package metadata (PyPI: vancam-mcp)
ā”œā”€ā”€ requirements.txt   # Python dependencies
└── .mcp.json          # Example MCP client config
```

## Related Projects

- [Vancam.ai](https://vancam.ai) — Web interface for traffic cameras
- [Model Context Protocol](https://modelcontextprotocol.io) — MCP specification
- [Vancam GPT](https://chatgpt.com/g/g-693512b18c0481918eb9b2c5d77e9eaa-vancam) — Same data, packaged as a ChatGPT GPT

## Contributing

Contributions are welcome — feel free to open an issue or submit a pull request.

## License

MIT — see [LICENSE](LICENSE).

TDQS

A4.7/5.0

Scored across 6 tools

Disambiguation5/5

Each tool targets a distinct spatial query mode (bounds, radius, route, nearest) plus image retrieval and API documentation. There is no overlap; an agent can easily select the right tool based on the user's query.

Naming Consistency5/5

All tool names follow a consistent verb_noun pattern in snake_case, with 'camera(s)' as the object. The verbs list, get, and describe are all clear and predictable, making the naming internally consistent.

Tool Count5/5

Six tools is an appropriate size for a camera discovery and retrieval server. Each spatial query type is represented, plus image fetching and a meta-documentation tool, with no unnecessary bloat.

Completeness5/5

The tool set covers all spatial search modes from the VanCam web app (bounding box, radius, route corridor, nearest neighbor) and includes live image retrieval that works around the API key limitation. There are no obvious gaps for the stated purpose.

Maintenance

ActivityStale
ResponsivenessNo issues