data-ai MCP Server
mysql-mcp — MCP server connecting Claude Code to MySQL. Supports SSH tunnels, SSL/TLS, and multiple databases. 20 tools for schema introspection, query execution, ERD generation, stored procedures, triggers, and events. Includes CLI for connection management.
Discovered via github-topic:mcp-server and last synced 3mo ago.
1. Install the package
npx querybridge-mcp
2. Add to claude_desktop_config.json
{
"mcpServers": {
"querybridge-mcp": {
"command": "npx",
"args": [
"querybridge-mcp"
]
}
}
}Config file location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) / %APPDATA%\Claude\claude_desktop_config.json (Windows)
Default
string
List all connections with status, host, SSH/SSL indicators
Mitigation
`querybridge-mcp`
Description
number
List tables with row counts and engine info
string
object
string
Show columns, indexes, and CREATE TABLE statement. Pass `table` for one table; pass `tables: [...]` for a batch (returns `{ results: [...] }`)
Run EXPLAIN in TRADITIONAL, JSON, or TREE format
Type
number
string
List views with definer, security type, updatability
Run SQL with parameterized values. Writes blocked on read-only connections
Get full DDL for a procedure or function
KILL QUERY (or KILL CONNECTION) by process ID. Gated: requires `readonly: false`
Search columns by pattern, sample tables, build a query
`stdio`
Opt out of bearer auth. Logged as a warning on every startup. Useful for `127.0.0.1`-only dev setups.
string
Switch the active database/schema for a connection
Show indexes with duplicate detection. `table` / `tables: [...]` / omit both for one / some / all
string
object or `true`
string
Show columns and CREATE VIEW DDL of a view
Stream a large SELECT to a NDJSON file on disk. Bypasses `execute_query`'s 1k-row in-memory cap. Row + byte caps both apply; hitting either issues `KILL QUERY` on the worker and marks the result `truncated`. Atomic rename on success; temp file is unlinked on failure
number
Behavior
Description
Get clean CREATE TABLE DDL. `table` for one, `tables: [...]` for a batch
List scheduled events with status and timing
string
List all databases accessible on a connection
Show FK relationships with cascade rules. `table` / `tables: [...]` / omit both for one / some / all
Preview rows from a table (default: 5 rows)
boolean
string
Find columns by name pattern across all tables
Discover tables, schemas, FKs, routines, triggers, events, and generate an ERD
Get clean CREATE VIEW DDL (raw, not truncated)
Row counts, data/index sizes, timestamps. Three modes: `table` for one, `tables: [...]` for a subset in one call, or omit both for every table (the "which tables are huge?" mode)
Get full event definition
Per-column profile — null %, distinct count, min/max/avg, optional top-N most common values. One combined-aggregation query per call; type-aware metric selection (no AVG on text, no MIN/MAX on BLOB)
Generate a Mermaid ER diagram with tables, columns, PKs, FKs, and relationships
Description
List stored procedures and functions
Show running connections + their current queries (filter by minimum duration)
Analyze a query with EXPLAIN, check indexes, suggest improvements
List triggers, optionally filtered by table
Detect secondary indexes with zero reads in `performance_schema` and produce DROP statements
Check for missing PKs, redundant indexes, empty tables, catalog routines and triggers
Get full trigger definition
Show character set and collation at database, table, and column levels
Diff a checked-in `.sql` schema file against a live database. Loads the file into a temp DB on a writable scratch connection, delegates to the same engine `compare_schemas` uses, and drops the temp DB. Designed for CI drift detection against a source-of-truth schema. V1 doesn't support `DELIMITER` blocks — stored routines / triggers in the file won't parse cleanly.
Diff two databases (potentially across connections). Reports drift across **9 aspects**: tables, table attributes (engine/charset/**partitioning**), columns (incl. comments, generated cols), indexes (incl. MySQL 8 invisible indexes, functional indexes, prefix lengths), foreign keys, views, routines, triggers, events. SQL bodies are whitespace-normalized; int display widths are normalized for cross-version (5.7 ↔ 8.0+) sanity. Restrict with `tables` filter or `scope` for cheaper runs. `summaryOnly: true` keeps huge diffs in context budget. Emits MCP progress notifications per scope. Honors client-side cancellation.
Databao agent is an open-source agent that enables you to chat with your data and receive answers in text, interactive charts, and tables.
Skardi is a self-improving context framework for ai agents
Lightweight task & matter service for the OCTO platform — REST API, timeline tracking, and LLM-powered extraction. Part of the OCTO open workplace for humans × AI agents.