Skip to main content
Meridian gives you two kinds of logs for a hosted app: build logs from turning your commit into an image, and runtime logs from the running service.

​
Runtime logs

Open Live logs from your app’s hosting overview. It shows everything the running container writes, the same output you would see in Cloud Run: standard output, the error stream, and every HTTP request with its method, path, status and latency. By default it tails the last 10 minutes of the current deployment and refreshes every few seconds. To investigate something that already happened, widen the window:
  • Range: the last 10 minutes, hour, 24 hours or 7 days, or further back up to your plan’s logs storage.
  • Level: everything, warnings only, or errors only. The filter is exact, so warnings and errors never appear in each other’s view.
  • Search: a word or a request path such as /webhooks/orders, matched before the line limit so a rare match is never pushed out by noise.
  • Current deployment only: on by default, so a redeploy does not mix in the previous container’s output. Switch it off to see the whole service, including earlier revisions.
Lines written to the container’s error stream carry a stderr tag and are shown as warnings even when the process gave them no severity, which is how most Node and Python loggers print errors. How far back the console reaches is your plan’s logs storage: 7 days on Lite, 30 days on Pro and Enterprise (Cloud Logging keeps no more than 30). Each view returns up to 500 lines. A quiet app fits comfortably; a busy one should narrow by level or search first.

​
Build logs

Every deployment links to the build logs for the image it produced. When a deploy fails in the building state, the build logs are where the compile or dependency error appears. You usually don’t need to open them to find it. A failed build shows Why it failed on the deployment: the error lines of the build step that failed (with the file, line and message when your compiler or bundler printed them), then Docker’s line naming the instruction that failed, last. When nothing in the output reads like an error, it shows the last lines of that step instead. The full output stays in the build logs.

​
Reading logs from your editor

Your AI client can tail runtime logs through Meridian MCP (logs_tail), scoped to an environment and optionally pinned to a single deployment’s revision, with severity and text filters, useful when debugging a rollout without leaving the editor.
Last modified on October 5, 2026