# MCP-wolfram-alpha

A MCP server to connect to wolfram alpha API.

<a href="https://glama.ai/mcp/servers/q5fud9cttp">
  <img width="380" height="200" src="https://glama.ai/mcp/servers/q5fud9cttp/badge" />
</a>

## Components

### Prompts

This is analogous to the `!wa` bang in duckduckgo search.

```python
def wa(query: str) -> f"Use wolfram alpha to answer the following question: {query}"
```

### Tools

Query Wolfram Alpha api.

```python
def query_wolfram_alpha(query: str) -> str
```

## Configuration

You **must** set the `WOLFRAM_API_KEY` environment variable. Get an api ket from [Wolfram Alpha](https://products.wolframalpha.com/api).

This was tested with the full results API, but it might not be required.

```json
{
    "mcpServers": {
        "MCP-wolfram-alpha": {
            "command": "uv",
            "args": [
                "--directory",
                "C:\\Users\\root\\Documents\\MCP-wolfram-alpha",
                "run",
                "MCP-wolfram-alpha"
            ],
            "env": {
                "WOLFRAM_API_KEY": "your-app-id"
            }
        }
    }
}
```

## Development

### Debugging

Since the official MCP inspector does not have good environment support, I reccommend using wong2's [mcp-cli-inspector](https://github.com/wong2/mcp-cli).

Create a config.json file in the same style as claude desktop.

```json
{
    "mcpServers": {
        "MCP-wolfram-alpha": {
            "command": "uv",
            "args": [
                "--directory",
                "/full/path/to/MCP-wolfram-alpha",
                "run",
                "MCP-wolfram-alpha"
            ],
            "env": {
                "WOLFRAM_API_KEY": "your-app-id"
            }
        }
    }
}
```

Then run:

```bash
npx @wong2/mcp-cli -c .\config.json
```