CommunityActive· 11d★ 4.3by designcomputer

MySQL

Read-only MySQL access for Claude. Schema-aware, runs SELECTs safely, perfect for ad-hoc analytics on a production replica.

View repo

Safety & Trust

Delv Safety Grade: C

Score 62/100 · assessed 2026-04-28

Maintainer45

Permissions75

Supply chain65

Transparency55

Incidents100

This community-built MySQL MCP server provides read-only database access through SELECT queries. The maintainer 'designcomputer' appears to be a solo developer with limited public profile, creating bus factor concerns. The read-only constraint is a strong safety feature, preventing data modification or deletion. However, the server requires full database credentials (host, user, password, database name) passed as environment variables, which means Claude gains access to potentially sensitive production data. The PyPI package distribution is standard, though the project lacks extensive documentation or community review. No security incidents are known, but the combination of database credential access and a single maintainer with unclear maintenance commitment warrants caution. Best used with dedicated read-only MySQL users on non-production replicas, never with credentials that have write privileges or access to sensitive customer data.

Lethal Trifecta (prompt-injection exposure)

ONE OF THREE

Private dataYes

Reads secrets, credentials, private files

Untrusted inputNo

Ingests web pages, PRs, issues, emails

External commsNo

Can send data outbound

Same.

Green flags

Enforces read-only access, preventing data modification or deletion
Schema-aware design reduces need to paste sensitive DDL into conversations
Distributed via standard PyPI package manager with versioning
Clear use case for analytics on production replicas
No known security incidents or malicious versions

Red flags

Solo maintainer with limited public profile and unclear maintenance commitment
Requires full database credentials including password in environment variables
Could expose sensitive production data if pointed at wrong database instance
Minimal documentation on security best practices or credential scoping
No evidence of security audit or peer review of query sanitization

Permissions requested

DB readOutbound networkAccess secrets

Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

Install

pip install mysql-mcp-server

Env vars needed: MYSQL_HOSTMYSQL_USERMYSQL_PASSWORDMYSQL_DATABASE

Review

This MCP server gives Claude read-only access to your MySQL database. Point it at a replica, feed it your credentials, and you can ask Claude to write and run SELECT queries against live schemas. It's schema-aware, so Claude knows your table structure without you pasting DDL into every conversation. I'd reach for this when I'm doing exploratory analytics on a dataset I don't know well yet. Instead of writing ten throwaway queries in a SQL client, I ask Claude to find patterns, aggregate metrics, or join tables I'd forgotten existed. It's particularly good for ad-hoc reporting tasks where the question changes every five minutes. The read-only constraint means you can point it at a production replica without the usual paranoia. Setup is straightforward if you're comfortable with environment variables. You'll need MYSQL_HOST, MYSQL_USER, MYSQL_PASSWORD, and MYSQL_DATABASE in your config. The server introspects your schema on connection, so Claude gets table names, column types, and relationships automatically. No manual schema dumps required. The main quirk is that it's read-only by design. If you need to run migrations, seed data, or test writes, this isn't the tool. That's a feature, not a bug, but it does mean you'll still need a traditional client for DDL work. Performance depends entirely on your MySQL instance. Complex queries on large tables will be slow because Claude is running actual SELECTs, not summarising cached metadata. Another thing to watch: Claude will sometimes write queries that are technically valid but inefficient. I've seen it do full table scans when an index hint would've been smarter. You're still responsible for checking the execution plan if performance matters. This is a community project, not an official Anthropic release. The repo is maintained by designcomputer and it works reliably, but don't expect enterprise support. If you're already running Claude Desktop and have a MySQL replica you want to query conversationally, this is a no-brainer. If you're on Postgres or need write access, look elsewhere.

Verdict

Install this if you have a MySQL replica and you're tired of writing throwaway SELECT queries by hand. Skip it if you need write access, work primarily in Postgres, or don't have a safe read-only instance to point it at. It does one thing well and stays out of the way.

Good at

Schema introspection is automatic. Claude knows your table structure without you pasting DDL.
Read-only by design, so you can point it at a production replica without risk.
Good for ad-hoc analytics where the question changes faster than you can write queries.
Works reliably with Claude Desktop and other supported hosts.

Watch out

No write access. You'll still need a traditional client for migrations or data seeding.
Claude sometimes writes inefficient queries that do full table scans instead of using indexes.
Community-maintained, so don't expect enterprise support or guaranteed updates.
Only works with MySQL. Postgres users need a different MCP server.

Getting started

1. Clone the repository from https://github.com/designcomputer/mysql_mcp_server and follow the setup instructions in the README to install dependencies. 2. Add the server to your Claude Desktop config file with the required environment variables: MYSQL_HOST, MYSQL_USER, MYSQL_PASSWORD, and MYSQL_DATABASE. 3. Restart Claude Desktop and verify the connection by asking Claude to list your database tables. 4. Test a simple SELECT query on a known table to confirm schema introspection is working. 5. Watch out for queries that run full table scans on large datasets. Claude won't always optimise for performance, so check execution plans if speed matters.

Works with

Claude DesktopClaude CodeCursor