MySQL (DBHub) for SQL databases
For SQL databases, Bytebase's DBHub pointed at MySQL is our second pick, a clean fit for the LAMP-era workhorse. The agent runs SQL and explores schema through one universal database server, which is exactly what turns a vague question into a concrete answer against MySQL.
It sits second because the Postgres build of DBHub leads for the most-used open engine, but MySQL still backs a vast install base, and this is the direct way to give an agent query access to it. DBHub is a community server rather than a vendor-official one.
How MySQL (DBHub) fits
Two tools cover the job. execute_sql runs queries with transaction support and safety controls, so an agent can explore data and, when permitted, apply changes inside a transaction. search_objects walks the schema (tables, columns, indexes, procedures) with progressive disclosure, which lets the agent default to reading and understanding before it writes, the safe pattern this task calls for.
The limit is the same minimal surface across all DBHub targets: no migration tooling and no distributed-SQL features, just SQL and schema search. Compared to the siblings, DBHub (Postgres) is the identical server aimed at Postgres and the lead pick there; Neon brings serverless Postgres with disposable branches per change; CockroachDB is the choice when you need a distributed SQL engine that survives node failures; ClickHouse fits columnar analytics rather than transactional work. When the engine is MySQL or MariaDB, DBHub is the right server.
Tools you would use
| Tool | What it does |
|---|---|
| execute_sql | Executes SQL queries against the connected database with transaction support and safety controls. |
| search_objects | Searches and explores database schema (tables, columns, indexes, procedures) with progressive disclosure to avoid overwhelming the context. |
FAQ
- Does DBHub default to read-only so an agent can explore safely?
- execute_sql runs with transaction support and safety controls, and search_objects lets the agent inspect schema before touching data, so the safe explore-then-mutate pattern is straightforward to enforce. Scope the connection's permissions to match how much you want the agent to be able to change.
- Should I use the MySQL or Postgres DBHub build?
- Point DBHub at the engine you actually run. The MySQL build fits MySQL and MariaDB; DBHub (Postgres) is the same universal server aimed at Postgres and leads for that engine. The tool surface (execute_sql, search_objects) is the same either way.