What reliable operations require
Monitoring starts with three questions: is the app reachable, are errors occurring, and how fast does it respond? We combine uptime checks, error tracking with tools like Sentry, and metrics from hosting and database. AI features add latency, error rate and token usage of model calls, because a provider outage or a sudden cost spike is just as much an incident as a server error.
Equally important are backups whose restore is rehearsed regularly, a runbook for common incidents and clear rules on who gets alerted and when. Without these basics, operations depend on the one person who happens to know the code. With AI-built apps that risk is higher, since the architecture is often undocumented, so we add concise operations documentation that coding agents can use as context too.
When something breaks at night
Every incident follows a fixed process, so nothing is left to chance. Up front we agree with you which systems are critical, who is informed at which severity and which actions we may take without asking first, so decisions do not stall in the middle of the night.
- A monitoring alert or your call to our 24/7 phone support opens a prioritized incident.
- The on-call engineers check monitoring and logs and restore service, by rollback or failover if needed.
- The root cause is analyzed with help from Claude Code or Codex and shipped as a tested fix through the pipeline.
- You receive a short post-incident review covering cause, impact and measures to prevent a repeat.
Ongoing maintenance for web and mobile
Operations also include planned work: security updates for dependencies, end-of-life runtimes, certificates and changes from AI providers such as deprecated models. For mobile apps, Google Play regularly requires a current target API level and Apple new SDK versions. We handle these store updates with appsubmitter.io and its automated pipelines in the customer's own account.
Logs and support tickets often contain personal data. Textwash Pro anonymizes them automatically before they are passed to coding agents or language models for analysis, an important building block for GDPR-compliant operations.
Frequently asked questions
Do you take over apps you did not build?
Yes. Before going on call we run an assessment, set up monitoring and fix the most important risks. Only then do we take on the on-call duty.
Where do hosting and data live?
In principle, in your own accounts with providers such as Vercel, Supabase, AWS or Hetzner. We receive the access we need, and you stay in control of infrastructure and data at all times.
How do we reach you in an emergency?
Through our 24/7 phone support in German and English. For agreed systems, monitoring alerts also go directly to the responsible team.