SQLite MCP Server
PaytonWebber/sqlite-mcp
Acesse SQLite via Claude com um binário estático de 1 MB
Ver no GitHubServidor MCP read-only para SQLite em um único binário de 1 MB, sem dependências de runtime. Permite que Claude consulte qualquer banco de dados SQLite com ferramentas de query, listagem de tabelas e acesso a schemas, com performance extremamente otimizada.
Instalação
npx convez-ai PaytonWebber/sqlite-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
- Consultar dados em bancos SQLite durante sessões Claude
- Analisar schemas de banco de dados sem expor operações de escrita
- Integrar Claude com aplicações que usam SQLite como backend
Veredito

Implementação limpa e extremamente eficiente de um MCP server para SQLite. A abordagem read-only é segura, o binário estático elimina fricção de deployment, e os benchmarks mostram vantagem clara sobre alternativas Python/Node. Ideal para quem quer acesso rápido a dados estruturados em Claude.
README
A tradução para o português deste README ainda está na fila.
sqlite-mcp
Point Claude at any SQLite database with a single 1 MB binary.
claude mcp add my-db /path/to/sqlite-mcp /path/to/database.db
No Python, no Node, no runtime. One statically linked executable with SQLite compiled in: about 1 MB stripped. A full cold-process session (initialize, handshake, one SELECT) completes in about 2 ms with 2.9 MB peak memory; the reference Python server takes about 410 ms and 64 MB for the same session, the most-used npm server about 210 ms and 84 MB.
The database is opened read-only. Writes fail at the SQLite layer, so it is safe to hand an LLM.
What it exposes
query tool | Run SQL, get rows back as pipe-separated text with a header. max_rows defaults to 100; truncation is reported. SQL errors come back as tool errors with the SQLite message. |
list_tables tool | Enumerate tables. |
sqlite://table/<name> resources | Each table's CREATE statement, so the model can read schemas before writing queries. |
Install
Download a binary from releases (Linux x86_64/aarch64 static, macOS x86_64/aarch64, Windows x86_64), or build from source:
zig build -Doptimize=ReleaseSmall # requires Zig 0.16.0
The build fetches the SQLite amalgamation and the zig-mcp-sdk through the Zig package manager; there are no system dependencies.
Claude Code
claude mcp add my-db /path/to/sqlite-mcp /path/to/database.db
Claude Desktop / Cursor
{
"mcpServers": {
"my-db": {
"command": "/path/to/sqlite-mcp",
"args": ["/path/to/database.db"]
}
}
}
Example session
> what tables are in this database?
list_tables -> orders, users
> who spent the most?
query("SELECT u.name, SUM(o.total) AS spent FROM users u
JOIN orders o ON o.user_id = u.id
GROUP BY u.id ORDER BY spent DESC LIMIT 1")
-> Alan Turing | 42.0
How it works
Built with zig-mcp-sdk. The tool schemas are generated at compile time from the argument structs, the SQLite amalgamation is compiled into the binary by Zig's C compiler, and every release is conformance-tested in CI: handshake, queries, NULL rendering, truncation, write rejection, schema resources.
License
MIT