MySQL AIops

Disclaimer: Community-maintained open-source project. Not affiliated with, endorsed by, or sponsored by Oracle Corporation or the MariaDB Foundation. "MySQL" is a trademark of Oracle Corporation; "MariaDB" is a trademark of MariaDB plc; all product/trademark names belong to their respective owners. MIT licensed.

Governed AI-ops for MySQL 8.x and MariaDB 10.6+ DBA operations — connecting to a server with PyMySQL and reading information_schema / performance_schema — with a built-in governance harness: unified audit log, token/runaway budget guard, undo-token recording, and descriptive risk-tier labels. The server flavor (mysql vs mariadb) is detected from version() and flavor-dependent statements branch automatically (SHOW REPLICA STATUS vs SHOW SLAVE STATUS).

What it does

Four flagship signature analyses, plus the guarded reads and writes around them:

• Slow-query RCA — take the worst events_statements_summary_by_digest entry (plus an optional EXPLAIN FORMAT=JSON plan) and map its numbers — no-index share (SUM_NO_INDEX_USED), lock-time share, rows-examined/sent ratio, tmp-disk spill, call count, plan access types — to a cited cause and a concrete action. Every finding carries its measured number, not a black-box verdict.
• InnoDB lock-wait & deadlock chain RCA — build the wait-for tree from performance_schema.data_lock_waits (MariaDB: information_schema.innodb_lock_waits), name the root blocker (blocks others, waits on none), and parse the last deadlock out of SHOW ENGINE INNODB STATUS.
• Replication lag RCA — map the replica's IO/SQL thread state, Seconds_Behind_Source and error fields to a cited cause + action (stopped IO thread, failed applier statement, lagging applier, intentional SQL_Delay).
• Table fragmentation analysis — rank tables by reclaimable data_free from information_schema.tables into cited OPTIMIZE TABLE candidates.

What works

• CLI (mysql-aiops ...): init, overview, server, activity, query, index, table, repl, analyze, remediate, secret, doctor, mcp.
• MCP server (mysql-aiops mcp or mysql-aiops-mcp): 35 tools (26 read, 9 write), every one wrapped with the bundled @governed_tool harness.
• Encrypted credentials: the account password lives in an encrypted store ~/.mysql-aiops/secrets.enc (Fernet + scrypt) — never plaintext on disk. Unlock with a master password from MYSQL_AIOPS_MASTER_PASSWORD (MCP/CI) or an interactive prompt (CLI).
• Reversibility: mutating writes fetch the real before-state first and record a faithful inverse — create_index↔drop_index; drop_index captures the index definition out of SHOW CREATE TABLE so undo recreates it exactly; set_global_variable captures the prior value from SHOW GLOBAL VARIABLES so undo sets it back. Irreversible ops (kill_session, kill_query, optimize_table, analyze_table, reset_query_stats) record prior state for audit but declare no undo.
• Safety: every state-changing CLI op supports --dry-run and requires double confirmation; every write MCP tool takes a dry_run preview. All identifiers that cannot be parameterised (schema/table/index/column/variable names) are validated against a strict charset and backtick-quoted; all values are bound query parameters.

What this tool does, and does not, decide

It delivers MySQL / MariaDB DBA operations — reads and writes — accurately and efficiently, and records every one of them. It does not decide whether a write is allowed to happen. That is the agent's judgement, or the permission of the account you connect it with: point it at a MySQL/MariaDB account granted only SELECT / PROCESS / REPLICATION CLIENT and no write privileges (no INSERT/UPDATE/DELETE/DDL), and the writes fail at the server — the place that actually owns the permission.

So there is no read-only switch, no policy file, no approval gate to configure. The one thing the tool guarantees is that nothing is silent: every call, over MCP and over the CLI alike, lands an audit row in ~/.mysql-aiops/audit.db, and destructive writes still capture their before-state and record an inverse where one exists.

Each tool declares a risk_level, carried into the audit row as a descriptive tier (none/confirm/review) — so a reviewer can see at a glance that a row was a high-risk delete. It is a label, not a gate.

Running a smaller / local model? See agent-guardrails.md — it lists the guardrails this tool now enforces for you (so you don't spend prompt budget restating them) and gives a ready-made system prompt for what's left.

Capability matrix (35 MCP tools)

The flagship analyses accept injected records for pure/offline analysis, or pull live from a configured target. top_queries/slow_query_rca require performance_schema=ON; the read account should have PROCESS, REPLICATION CLIENT and SELECT on performance_schema.

Support scope

Quick start

As a Claude Code plugin

One install gives an agent both the skill and the MCP server:

/plugin marketplace add AIops-tools/marketplace
/plugin install mysql-aiops@aiops-tools

The MCP server is fetched with uv and pinned to the package version this plugin declares, so an audit row can be traced back to the code that wrote it. Credentials are still configured with mysql-aiops init — see below.

As an OpenClaw plugin

The same bundle is published on ClawHub, where one install delivers the skill and its MCP server together:

openclaw plugins install clawhub:@zw008/mysql-aiops
openclaw skills info mysql-aiops          # expect: Visible to model: yes

Restart the OpenClaw gateway afterwards so it loads the plugin. The MCP server is fetched with uv, pinned to this exact release, so uvx has to be on PATH — without it the skill still installs but reports Visible to model: no. Credentials are configured exactly as below.

As a CLI or standalone MCP server

uv tool install mysql-aiops             # or: pipx install mysql-aiops
mysql-aiops init                        # wizard: add a target + store the password (encrypted)
mysql-aiops doctor                      # verify config, secrets, connectivity, flavor, perf-schema
mysql-aiops overview                    # one-shot server health snapshot
mysql-aiops analyze slow-query          # RCA the worst statement digest
mysql-aiops analyze fragmentation       # OPTIMIZE TABLE candidates

Run as an MCP server (stdio):

export MYSQL_AIOPS_MASTER_PASSWORD=...  # unlock secrets non-interactively
mysql-aiops-mcp

Where that password then lives: an exported variable is readable by every process this shell starts and is recorded by shell history. On a shared or long-lived host, prefer the interactive prompt, or inject it from a secret manager for the life of the one command that needs it.

Claude Desktop / MCP client config:

{
  "mcpServers": {
    "mysql-aiops": {
      "command": "uvx",
      "args": ["--from", "mysql-aiops", "mysql-aiops-mcp"],
      "env": { "MYSQL_AIOPS_MASTER_PASSWORD": "your-master-password" }
    }
  }
}

Env-block caveat: the env block above is the only environment the MCP server sees — GUI-launched clients do not inherit your shell profile. Put MYSQL_AIOPS_MASTER_PASSWORD (and MYSQL_AIOPS_HOME / MYSQL_AUDIT_APPROVED_BY if you use them) there, or the server cannot unlock the secret store.

Governance

Every MCP tool passes through the bundled @governed_tool harness: