Log Central
A fleet-wide observability console — logs, error groups and API traffic for every registered app in one place, with client packages and an MCP server so even the AI tooling can read it.
When you operate many applications — payment services, mail platforms, client portals — “check the logs” stops being a command and becomes a scavenger hunt. Log Central is the answer I built: one console where every registered app ships its logs, exceptions and API request traffic.
The shape of it
- Ingestion clients, not agents. Apps opt in with a package — Laravel, Node, or a ZoneMTA plugin for the mail layer — that ships structured events to the console. Adding an app to the fleet is a composer/npm install, not an infrastructure project.
- Error groups with a lifecycle. Exceptions are grouped by fingerprint; a group can be resolved, snoozed, or reopened, and carries incident notes. The unit of attention is “this problem”, not “these 4,000 identical lines”.
- Traces across the seams. An error carries the request input that caused it and a trace id; the same trace id replays the surrounding log lines. The debugging flow — find the group, read the stack, replay the context — is one console, three clicks.
- API health as a first-class view. Per-endpoint stats, per-client traffic, error rates — the questions you ask when a partner says “your API is slow” have a page.
- An MCP server on top. The console exposes its search and stats as Model Context Protocol tools, so an AI assistant can run the same debugging flow I do — list apps, find the failing group, pull the trace — without screen-scraping.
The opinion embedded in it
Observability tools drown you by default. Log Central is add/edit only in its management plane (deletion demands the web console deliberately), timestamps are UTC everywhere, and the fleet view is the default view — because the question is never “is app X fine?” but “what, anywhere, is on fire?”