Last week, php2shell turned a routine PHP update into an urgent race across 50+ servers.
Our managed servers normally update nightly. This time, waiting wasn’t an option.
To make things worse, update servers were overloaded. Downloads failed, updates stalled, and every server needed monitoring.
So we put an AI agent to work through our server-management service via MCP.
Its job was simple:
→ Find servers running an old PHP version
→ Detect failed updates
→ Retry until the update succeeded
No manually babysitting 50+ servers. Just a persistent agent working through the failures.
AI is accelerating incidents. We should use it to accelerate our response, too.
commit 557ef73