Your agent can suggest improvements, but it can’t see what happened yesterday. DBMetrics collects query and workload history and makes it available to your agent.
How it works
You ask Claude“Why is my database slow?”
Via MCP, the agent sees latency jumped after a migration 2 days ago.
A new query needs an index. Claude adds it with your approval.
Fresh metrics confirm the same query is faster.
Collect metrics from different databases at once. Your agent reads the history through DBMetrics to investigate slowdowns, optimize queries and verify improvements.
Connect through MCP, API or CLI.
Multiple databases, connected together
PostgreSQLExtension
MySQL
MongoDBNo inbound database access
Database metrics, stored over time
Run on your server or in the cloud
Docker
UbuntuCloudClaude Code, Codex or OpenCode
Run the server, connect your database, then give your agent access. Follow the commands or copy a prompt and let your agent help.
Setup preview: packages and endpoints below are illustrative.
Store your metrics on infrastructure you control.
TERMINAL
Ask your agentSend database metrics to your DBMetrics server.
Use a dedicated monitoring user.
SQL + TERMINALAsk your agentGive your agent access to the metrics history through MCP.
Your agent can read the history from every connected database.
Ask your agentYou run the collector. DBMetrics Cloud handles storage and updates. Cloud signup is coming soon.
Run DBMetrics for free on your infrastructure or pay for managed storage in the cloud. Connect your existing agent with either option.
Your server. You manage storage and updates.
$0software cost
any number of databases
You cover your own infrastructure costs.
You run the collector. We host the metrics server.
$19per database
per month · USD
What counts as a database? One monitored database connection. A replica counts separately if you connect it for monitoring.
Same price for every database type. Your agent subscription is paid separately.
Yes. Direct access can show its current state. DBMetrics continuously records query and workload history so your agent can also investigate what happened yesterday, during a traffic spike or before a release.
DBMetrics collects observations and makes them available to your agent. It doesn’t apply indexes or change configuration. You review recommendations and use your existing workflow to make and verify changes.
Query frequency and timings, schema and indexes and available workload metrics such as connections and locks. The exact set depends on the database and how it’s deployed. Query text can contain sensitive data; review collection settings before connecting production.
The collector can run near a database and connect with a monitoring user. Managed providers differ in the statistics and extensions they expose, so the available metrics depend on your provider and permissions.
Use your existing Claude Code, Codex or OpenCode setup. DBMetrics supplies the metrics through MCP. Your agent subscription and model usage stay with your existing provider.
SQLite runs inside your application. Collect query timings with an application-side integration and send them to DBMetrics. An external collector can inspect the database file but cannot reconstruct timings for queries your app already ran.