Sakenowa MCP Server
yawaragi-dev/sakenowa-mcp
Consultas estruturadas sobre sake japonês via MCP para Claude
Ver no GitHubServidor 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-mcp1 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
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
| Variable | Required? | Default | Purpose |
|---|---|---|---|
DATABASE_URL | yes | — | Postgres connection string (read-only role recommended) |
MCP_LOG_LEVEL | no | error | silent / 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.