# MCP Research Server

AI-powered web research tools for Claude Desktop, Cursor, and any MCP-compatible client.

**Works with OpenRouter (free) or Anthropic API.**

## Tools

### `research_company`
Deep company intelligence — website analysis + AI synthesis.

```
What does Linear do and who are their competitors?
→ Uses research_company("Linear", url="linear.app")
```

Returns: description, industry, business model, size estimate, products, pricing model, competitors, strengths.

### `summarize_url`
Fetch and summarize any URL in concise/detailed/bullets format.

```
Summarize this blog post: https://...
→ Uses summarize_url(url, style="bullets")
```

### `compare_companies`
Side-by-side comparison of 2-3 companies.

```
Compare Notion vs Linear vs Asana
→ Uses compare_companies(["Notion", "Linear", "Asana"])
```

Returns: markdown comparison table + conclusion.

### `extract_data`
Extract specific structured data from any webpage.

```
Extract the pricing plans from https://vercel.com/pricing
→ Uses extract_data(url, extraction_goal="pricing plans")
```

Returns: JSON of extracted data.

## Install

```bash
pip install openai mcp httpx beautifulsoup4
```

## Claude Desktop Config

### Option A: OpenRouter (Free — recommended)

```json
{
  "mcpServers": {
    "research": {
      "command": "python3",
      "args": ["/path/to/server.py"],
      "env": {
        "OPENROUTER_API_KEY": "your-openrouter-key"
      }
    }
  }
}
```

Get a free OpenRouter key at [openrouter.ai](https://openrouter.ai) — no credit card required for free models.

### Option B: Anthropic API

```json
{
  "mcpServers": {
    "research": {
      "command": "python3",
      "args": ["/path/to/server.py"],
      "env": {
        "ANTHROPIC_API_KEY": "your-key-here"
      }
    }
  }
}
```

## Cost

- **OpenRouter free models**: $0.00/call — completely free
- **Anthropic Haiku**: ~$0.001-0.005/call

## Requirements

- Python 3.10+
- `OPENROUTER_API_KEY` (free) or `ANTHROPIC_API_KEY`

## License

MIT