convez.ai
MCP ServerIntegração & APIsIntermediário
yawaragi-dev

Sakenowa MCP Server

yawaragi-dev/sakenowa-mcp

Consultas estruturadas sobre sake japonês via MCP para Claude

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que expõe dados de sake japonês de um espelho PostgreSQL do Sakenowa como ferramentas Claude. Fornece busca por nome, recomendações por similaridade, filtros por prefeitura e rankings de popularidade — tudo read-only e stateless.

Instalação

$npx convez-ai yawaragi-dev/sakenowa-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
50%
Último commit
há 4 meses
Versão
v0.1.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Recomendações de sake baseadas em perfil de sabor
  • Busca e detalhes de sakes específicos por nome
  • Filtros por prefeitura, tags e características de sabor
  • Análise de popularidade e rankings regionais de sake

Veredito

MCP server bem arquitetado para domínio específico (sake), com schema limpo e ferramentas focadas. Projeto em v0.1.0 (ativo, mas ainda em desenvolvimento) — útil se você já tem Sakenowa espelhado em Postgres e quer acesso estruturado via Claude.

README

A tradução para o português deste README ainda está na fila.

@yawaragi/sakenowa-mcp

A Model Context Protocol server that turns well-defined queries over a Postgres mirror of Sakenowa sake data into MCP tool calls.

Connect it to Claude Desktop, the AI SDK's MCP client, or any other MCP consumer, point it at a Postgres that has a Sakenowa-mirrored schema, and the consumer can ask for sake recommendations, similarity search, prefecture filters, and the latest popularity rankings.

The server is read-only, stateless, and intentionally domain-pure — no LLM calls, no cross-beverage heuristics, no user identity, no business logic. The OSS asset's value is its independence: any consumer that mirrors Sakenowa data into Postgres can plug this in.

Status: v0.1.0 in development. The current scope is captured in docs/specs/v0.1.0.md. Open issues on GitHub track the implementation slices. Star the repo to follow progress.

What's in v0.1.0

Six MCP tools over the Sakenowa data shape:

  • list_prefectures — Japan's 47 prefectures.
  • search_sakes_by_name — resolve a free-text name (romaji or kanji) to one or more Sake records.
  • get_sake_details — full record for one Sake, including its 6-axis flavor profile, tags, brewery, and prefecture.
  • find_similar_sakes — cosine similarity over the 6-axis flavor profile.
  • find_sakes_by_flavor — filter by axis ranges, tag membership, and / or prefecture.
  • get_top_ranked — latest overall or per-prefecture popularity ranking.

Full input / output shapes and SQL intent: docs/specs/v0.1.0.md.

Use

pnpm add -g @yawaragi/sakenowa-mcp   # or npm i -g / yarn global add

The server reads DATABASE_URL from the environment and serves MCP over stdio.

Claude Desktop config

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on your platform:

{
  "mcpServers": {
    "sakenowa": {
      "command": "npx",
      "args": ["-y", "@yawaragi/sakenowa-mcp"],
      "env": {
        "DATABASE_URL": "postgresql://user:pass@host:5432/your_sakenowa_mirror"
      }
    }
  }
}

Restart Claude Desktop. The six tools become available in any chat.

AI SDK (Vercel) consumers

import { experimental_createMCPClient as createMCPClient } from 'ai';
import { Experimental_StdioMCPTransport } from 'ai/mcp-stdio';

const mcp = await createMCPClient({
  transport: new Experimental_StdioMCPTransport({
    command: 'npx',
    args: ['-y', '@yawaragi/sakenowa-mcp'],
    env: { DATABASE_URL: process.env.SAKENOWA_DB_URL! },
  }),
});

const tools = await mcp.tools();
// pass `tools` to generateText / generateObject / streamText

Getting a Sakenowa-mirrored Postgres

This server reads from Postgres; it does not fetch from Sakenowa's API at query time and does not ship an ingest pipeline. You bring the data.

The expected schema is documented in docs/specs/v0.1.0.md. Any ingest that produces tables matching that shape works.

Configuration

VariableRequired?DefaultPurpose
DATABASE_URLyes—Postgres connection string (read-only role recommended)
MCP_LOG_LEVELnoerrorsilent / error / info / debug. Anything other than silent writes to stderr only.

Attribution

Sake data exposed through this server comes from Sakenowa. Their data is free under an attribution-only licence. Any application or surface that displays this data to users must render the appropriate Sakenowa attribution. "Flavor Chart" is a registered trademark of Sakenowa; refer to the 6-axis visualisation as "flavor chart (Sakenowa)" on first mention.

Attribution is the consumer's responsibility — the MCP server returns the data; the application using the server renders the attribution.

Contributing

The implementation is being built slice-by-slice; the GitHub issues track open work, each labelled ready-for-agent once fully specified. Read AGENTS.md for the project conventions before opening a PR.

License

MIT. See LICENSE.

Ver o original no GitHub

Bom para

Relacionados