Provides read-only tools for inspecting and querying a SQLite database, including schema inspection and parameterized SELECT queries with pagination.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Shop Database MCP Server Show me the top 5 products by total revenue."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
This local stdio MCP server lets an AI agent inspect and analyze shop.db. It is read-only: SQLite is opened read-only, query_only is enabled, and the server rejects anything except one SELECT or WITH ... SELECT query.
Node.js 20 or newer and the supplied SQLite database.
Related MCP server: shop-database-mcp
npm installSHOP_DB_PATH is optional. If set, a relative value is resolved from the process working directory; an absolute value is used directly. Otherwise the server finds shop.db relative to its compiled module, including when launched outside the project directory. See .env.example; it is not loaded automatically.
npm run buildnpm startA stdio server normally appears idle because it waits for MCP messages.
In the IDE extension: Gear menu → MCP servers → Add server → STDIO. Set command, arguments, and environment from config/codex.config.toml.example, save, and restart the extension.
For a project-scoped setup, copy that example to .codex/config.toml and replace every /ABSOLUTE/PATH/TO/PROJECT placeholder. Codex CLI and the IDE extension share Codex MCP configuration. Do not commit the active configuration file.
inspect_database takes no parameters. Use it before querying or when uncertain; it returns tables, columns, primary/foreign keys, relationships, and row counts.
query_database accepts sql (required), params (positional string, finite number, boolean, or null values), limit (1–200; default 100), and offset (default 0). Use ? placeholders; joins and aggregates work. It accepts exactly one read-only SELECT or WITH ... SELECT. Inspect schema before assuming columns, date formats, prices, or statuses. Results contain columns, rows, rowCount, limit, offset, and hasMore; request another page with a new offset.
{"sql":"SELECT country, COUNT(*) AS customer_count FROM customers GROUP BY country ORDER BY customer_count DESC","params":[],"limit":100,"offset":0}npm testTests create temporary fixture databases and never modify shop.db.
Safety layers are read-only file access, SQLite query_only, conservative token-aware validation that ignores comments/strings/quoted identifiers, a single-statement rule, and server-side pagination. Tool errors are concise and do not expose stack traces or local paths.
If startup fails, verify SHOP_DB_PATH names an existing regular SQLite file. If a query fails, call inspect_database, use ? parameters, and verify names and relationships.
Use before your first query or whenever the schema is uncertain. Returns tables, columns, keys, relationships, and row counts.
Inputs: no named parameters. Returns: tables
Execute one parameterized read-only SELECT or WITH ... SELECT statement. Use ? placeholders and params for dynamic values. Joins and aggregate queries are supported; schema and data changes are forbidden. Results are paginated: use offset for the next page when hasMore is true. Inspect the schema instead of guessing column names, date formats, price fields, or order status semantics.
Inputs: sql (required), limit, offset, params. Returns: rows, limit, offset, columns, +2 more
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 2 tools
inspect_database handles schema metadata while query_database executes read-only SQL queries. There is no functional overlap or ambiguity between the two tools.
Both tools follow the same lower_snake_case verb_noun pattern: inspect_database and query_database. The naming is consistent, predictable, and clearly reflects each tool's purpose.
Two tools is lean but well-suited to the server's read-only database purpose: inspect the schema, then query the data. Each tool is essential, and the narrow scope justifies the small count.
For a read-only database server, inspect_database plus query_database covers the complete workflow of schema discovery and arbitrary SELECT queries with pagination. Write operations are explicitly out of scope, so their absence is not a gap.
Explore, query, and inspect SQLite databases with ease. List tables, preview results, and view det…
PostgreSQL, MySQL, and SQL Server in one session. 26 read-only MCP tools for AI agents.
Safe, read-only Postgres and MySQL access for AI agents. Audit log + column-level controls.
Query, explore, and manage your databases through natural conversation.
