See what your customers see, before they call
MonX is service monitoring for banks that starts from the customer's side of a service. I co-founded it in November 2024 and built the first version end to end.
Green dashboards, failing payments
Most monitoring watches infrastructure: servers, databases, the network. Every light can be green while card payments quietly fail, and the bank hears about it from customers first. Regulators now expect banks to classify and report those incidents against a deadline, so a missed outage costs paperwork as well as trust.
Dashboards can be green while customers can't pay.
MonX watches business outcomes instead of machines: card payments, logins and transfers per minute. It learns what normal looks like for every hour and kind of day, flags unusual drops, shows which other signals moved at the same time and drafts the incident reports banks owe their regulator.
- 01
Agents measure what matters
Scheduled SQL, PowerShell and REST checks run inside the bank's network and turn their results into indicators: payments per minute, failed logins, transfer times.
- 02
It learns what's normal
Every indicator gets its own baseline for each hour across ten kinds of day, including holidays and the first and last working day of the month, so a quiet Sunday doesn't look like an outage.
- 03
It explains and reports
When something drifts, MonX alerts the right channel, ranks the signals that moved with it and keeps an incident register with regulatory deadlines and draft reports.
Inside the product
- Anomaly detection on business indicators, not just infrastructure
- "What moved together": the ten signals most related to a drop
- Alerts to email, SMS, Teams, Slack, Jira, ServiceNow and more
- Incident register aligned with Serbian central bank rules and DORA
- Reads tools banks already run, such as Prometheus, Zabbix and Dynatrace
- Optional AI assistant that can run on the bank's own model
Where it stands
- Presented at the Bank Info conference in November 2025
- Demoed live to banks and payment companies in the region
- Preparing its first bank release
Role: Co-founder · product and development
Engineering notes
Stack
- .NET 10
- SQL Server
- React 19
- PowerShell
- On-prem
- 01
Baselines use the median and median absolute deviation per hour and day type, chosen over a neural model after testing both.
- 02
Correlation runs on each signal's deviation from its own baseline, so two metrics don't look related just because both follow the time of day.
- 03
Secrets are write-only: encrypted with AES-256-GCM under RSA-wrapped keys that only the agent service can open.
- 04
The web side and the agent side never call each other. They meet only in the database.
Want to talk about MonX?
A partnership, an investment, a pilot or just a question. Write, and I'll reply personally.
Get in touch
