MCP Database Server
One MCP server for SQLite, SQL Server, PostgreSQL and MySQL: read and change data, edit schemas and export to CSV and JSON
- 390 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The agent changes data and schema: UPDATE, DELETE, ALTER TABLE and dropping tables are available, and there is no read-only mode
- The database password is passed as a launch argument and kept in the client configuration as plain text
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add mcp/mcp-database-serverDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent changes data and schema: UPDATE, DELETE, ALTER TABLE and dropping tables are available, and there is no read-only mode
- The database password is passed as a launch argument and kept in the client configuration as plain text
This is third-party code. Review the repository files before installing.
What it does
MCP Database Server gives an agent SQL access to four kinds of databases: SQLite, Microsoft SQL Server, PostgreSQL and MySQL. The launch argument picks the database type. For SQLite a file path is enough; for the others you give a server or host, a database name and, if needed, a login and password. Amazon RDS MySQL supports IAM sign-in with SSL turned on automatically. The server has ten tools. `read_query` runs SELECT and `write_query` takes INSERT, UPDATE and DELETE. `create_table` and `alter_table` change the schema, and `drop_table` removes a table only with explicit confirmation, `confirm: true`. `list_tables` and `describe_table` show how the database is built, and `export_query` writes a query result to CSV or JSON. The pair `append_insight` and `list_insights` keeps notes with conclusions about the data so the agent can collect them during analysis. The server starts through npx from the `@executeautomation/database-server` package and is connected in the Claude Desktop configuration as a separate entry for each database. For different databases and credentials you can keep several instances side by side.
Who it is for. For developers and analysts who want to ask questions of SQLite, PostgreSQL, MySQL or SQL Server from Claude Desktop and export results to files.
Good fit when
- You need to explore an unfamiliar database: the agent lists tables, reads their structure and explains the links between them
- You want to pull a sample quickly from SQLite, PostgreSQL, MySQL or SQL Server and export it to CSV or JSON
- You do analytics and want the agent to record the patterns it finds as separate notes with conclusions
- The project has several different databases and each needs its own instance with its own credentials
- You need to connect to MySQL on Amazon RDS through IAM, with no password in the configuration
Not a fit when
- A database with production data and no backup: the server allows changes and deletion and has no read-only mode
- You need other systems such as Oracle, MongoDB or ClickHouse: the server supports only the four listed database types
- You want hints on indexes, query plans and database health: the tool set has no such analysis
- You need a shared server for a team: the process starts locally on your machine over stdio and has no network mode
Example request
List the tables in the database, describe the structure of orders and customers and export to CSV the ten customers with the largest order totals.MCP
- Transport
- stdio
- Authentication
- API key
Limitations
The server has no read-only mode: `write_query`, `alter_table` and `drop_table` are always available, so limit permissions on the database side with an account that has minimal privileges. The password goes in the launch arguments and sits in the client configuration as plain text. MySQL on Amazon RDS supports IAM sign-in, which needs configured AWS credentials.
Node.js 18 or newer is required, SQL Server 2012 or later and PostgreSQL 9.5 or later. If you give no login for SQL Server, Windows authentication is used. The latest commit in the repository dates from August 2025.
The server talks only to the database you name. Availability from Russia depends on where your database lives.
How to disable. Delete the entries with @executeautomation/database-server from claude_desktop_config.json and restart Claude Desktop. If you installed the package globally, run `npm uninstall -g @executeautomation/database-server`. Also revoke the password of the database account that was in the configuration.
FAQ
How do I connect a database to Claude Desktop through this server?
Add an entry to the mcpServers section of claude_desktop_config.json with the command npx and the arguments -y, @executeautomation/database-server and the database parameters. For SQLite a file path is enough; the other types need a host or server and a database name.
Which databases does the server support?
SQLite, Microsoft SQL Server, PostgreSQL and MySQL, including MySQL on Amazon RDS with IAM sign-in. The minimum versions are SQL Server 2012, PostgreSQL 9.5 and Node.js 18.
Can the agent delete data?
Yes. The `write_query` tool takes INSERT, UPDATE and DELETE, and `drop_table` removes a table after the confirmation `confirm: true`. The server cannot work read-only, so give it an account with minimal rights.
How do I export a query result to a file?
Ask the agent to export the sample: the `export_query` tool takes a SELECT and a format, CSV or JSON.
What are append_insight and list_insights for?
They are notes with business conclusions. The agent adds an observation with `append_insight`, and `list_insights` shows all the notes collected so conclusions do not get lost between analysis steps.
Discussion
No questions or reviews yet.
Related
MCP serversA single-binary MCP server for MySQL, MariaDB, PostgreSQL and SQLite: read-only by default, with optional PII redaction
A Python MCP server for MySQL: listing tables, reading data, running SQL and inspecting schema through a controlled interface
MCP server that gives an agent access to MySQL: schema inspection and queries, read-only by default, writes behind flags
An MCP server for Microsoft SQL Server and Azure SQL: the agent lists tables and runs SQL queries that read and write data
MCP SQLite Server (npx)
MCP SQLite Server
Node.js MCP server for SQLite databases that runs via npx, no Python or uvx needed
A desktop SQL workspace with a built-in MCP server: the agent reads the schema and runs queries in the same client you already use
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| MCP Database Server (this card) | no | API key | 390 | |
| MCP Toolbox for Databases | yes | not required | 16,812 | Google's set with ready tools for Postgres, MySQL, BigQuery and other databases |
| MCP SQLite Server (npx) | no | not required | 17 | a SQLite-only server that runs through npx |
README in short
The README describes two ways to run: the ready npm package and a local build from source. For each database it gives the command and the lists of required and optional parameters, then examples of Claude Desktop configuration blocks for both ways and a table of the ten tools. Detailed SQL Server and PostgreSQL guides and sample queries live in separate documents in the docs folder.