hvilkenai-mcp

hvilkenAI MCP Server


Daily Scandinavian AI language quality benchmark — Norwegian, Swedish, Danish.

What this does

hvilkenAI benchmarks AI models daily on practical Scandinavian language tasks. 12+ models are tested every day, selected from a pool of over 350 AI models available through OpenRouter. This MCP server lets AI assistants access fresh benchmark data directly.

Use cases:

• "Which AI is best at Norwegian right now?"
• "Compare GPT-4o vs Claude on Swedish"
• "What's the cheapest model that handles Danish well?"
• "Show me how Gemini has performed this week"

Available tools

get_daily_benchmark

Get today's benchmark results for AI models on Scandinavian languages.

| Parameter | Type | Values | Default | |-----------|------|--------|---------| | language | string | no, sv, da | no | | tier | string | all, premium, midrange, budget | all |

Returns: Model rankings with scores, pricing, and speed. Free tier returns yesterday's top 3; Pro returns today's full list.

get_model_history

Track a specific model's performance over time.

| Parameter | Type | Values | Default | |-----------|------|--------|---------| | model_name | string | Model name or partial match, e.g. "Claude Sonnet", "GPT-4o", "gemini" | (required) | | language | string | no, sv, da | no | | days | number | 1–30 (free tier: max 3) | 7 |

Returns: Daily scores and trends for the specified model.

get_orchestrator_ranking

Find the best model for AI agent orchestration and multi-agent systems.

| Parameter | Type | Values | Default | |-----------|------|--------|---------| | language | string | no, sv, da | no |

Returns: Models ranked by combined language quality and instruction-following ability.

get_recommendation

Get a model recommendation tailored to your specific needs and budget.

| Parameter | Type | Values | Default | |-----------|------|--------|---------| | use_case | string | writing, coding, research, agent, general | general | | language | string | no, sv, da | no | | budget | string | free, cheap, any | any |

Returns: Best model match with reasoning. Pro tier returns top 3 with full details.

get_weekly_summary

Get the weekly benchmark summary with trends and reliability analysis.

| Parameter | Type | Values | Default | |-----------|------|--------|---------| | language | string | no, sv, da | no | | week | string | Format YYYY-WW, e.g. "2026-21" | current week |

Returns: Winner, most reliable model, and key observations for the week.

Quick start

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "hvilkenai": {
      "type": "streamablehttp",
      "url": "https://mcp.hvilkenai.no/mcp"
    }
  }
}

Claude Desktop supports OAuth 2.1 — you will be prompted to connect with free or Pro access when you first use the server.

Claude Code (CLI)

claude mcp add --transport http hvilkenai https://mcp.hvilkenai.no/mcp

Cursor / VS Code

{
  "mcpServers": {
    "hvilkenai": {
      "url": "https://mcp.hvilkenai.no/mcp"
    }
  }
}

Direct API

curl -X POST https://mcp.hvilkenai.no/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'

Authentication

Free tier — no API key required:

• 24-hour delayed data
• Top 3 models per language
• 100 requests/hour

Pro — requires API key (Bearer token):

• Real-time data (updated daily at 07:30 CET)
• All models, all languages
• Historical data up to 30 days
• Higher rate limits

Get an API key at hvilkenai.no/priser/

For Pro access, add your API key as a Bearer token:

{
  "mcpServers": {
    "hvilkenai": {
      "type": "streamablehttp",
      "url": "https://mcp.hvilkenai.no/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Example interaction

You: Which AI model is best at Norwegian today? Claude (using hvilkenAI): Based on today's benchmark from hvilkenai.no, [model name] leads with a score of X.X, followed by...

About the benchmark

• Updated daily at 07:30 CET
• Three languages: Norwegian (Bokmål), Swedish, Danish
• Practical tasks: Real-world language use, not academic tests
• Broad coverage: 12+ models selected daily from over 350 AI models
• Independent: No sponsorships, no affiliate deals
• Proprietary methodology: Scoring uses a proprietary weighting system

Data source: hvilkenai.no

Links

• Website: hvilkenai.no
• Glama: glama.ai/mcp/servers/erorund/hvilkenai-mcp
• Integrations page: hvilkenai.no/integrasjoner/
• Server URL: https://mcp.hvilkenai.no/mcp
• Health check: https://mcp.hvilkenai.no/health

License

MIT