From your table structure and real data, AI works out what service the database runs and where the core of the business lies. Instead of CPU or connection counts, you get the numbers that matter to your business, like order volume and payment success.
DB
Watches successful and failed payments together to catch failures that cluster.
SELECT count(*) FILTER (WHERE status = 'failed') AS failed, count(*) FILTER (WHERE status = 'succeeded') AS succeeded FROM payments WHERE created_at >= now() - interval '5 minutes'
No time to work out what you should be monitoring
You watch CPU and connections, but wouldn't notice if orders stopped
Only the person who built a monitor knows what it means
AI reads your schema and real data, decides which tables are central to the business, and writes monitors for them. It also records what each database is for.
Each numeric column a monitor query returns becomes its own series. Values best judged together, like successful and failed payments, can be watched as a set.
Each monitor comes with a description of what it's meant to catch. The chart makes sense even when the person who built it isn't around.
During patrols, AI reviews the monitors and the database descriptions. Monitors no longer needed are deleted, and those whose intervals no longer fit are adjusted.
Monitors written by AI appear as charts. You can inspect everything about them on screen.
A single SELECT statement that returns one row. It runs over a read-only connection, and only after passing the SQL guard.
AI picks from line, bar, stacked, and pie charts. Values with different units or scales go on the right-hand axis.
Values are collected at intervals from 30 seconds to 1 day, and AI judges them on a separate interval. The shortest judgment interval depends on your plan.
AI's understanding of what system the database belongs to and which tables are central to the business is saved for each connection.
Collection and judgment can be paused per monitor. You can also pause the whole project.
AI deletes monitors that have served their purpose. You can also delete them by hand from the console.
All you need is a database connection string or an account connection. Nothing to install on your servers.
Register a connection
Register a connection string. If it has write privileges, a read-only user for monitoring is created on the spot. For Cloudflare D1 and BigQuery, connect your account and pick the database or project.
AI builds the monitors
AI reads the schema and real data, creates monitors, and lays them out as charts.
Check the charts
See how each monitor's values change over time, and when AI looked and what it decided.
E-commerce
Watch order volume and payment success
After reading the order and payment tables, AI creates monitors for recent order volume and for successful and failed payments. It catches changes like orders stopping or failures piling up.
SaaS
Watch the core actions of your service
The key numbers, such as posts or sign-ups, differ from service to service. AI reads them from your table structure and builds monitoring that fits your service.
Anomaly detection
AI judges whether something is abnormal based on value trends, past alerts, and your project's memory.
Patrols
Even on days with no anomalies, AI reviews monitors, database descriptions, and memory, and maintains only what needs it.
Project memory
Facts found in investigations and business context are recorded and used in every later judgment and investigation.
Leave what to watch
to AI.
Register a single connection string and Scandock starts by designing your monitors.