Glitches

Edge-observed faults on your own project's traffic: missing buckets and objects, dead routes, and 5xx from your app, grouped and ranked. Report-only, metadata-only, on by default.

Overview

Glitches (the glitch tracker) surfaces the faults the Scalix edge already observed on your project's traffic while serving it: storage reads that named a missing bucket or object, dead routes, broken auth, rate limits, and 5xx from your own app. Groups are ranked by frequency with first/last seen and a bounded sample of real paths, so you learn what your deployment is getting wrong without filing a support ticket.

The feed is on by default for every project. There is nothing to enable to see it. It is:

Report-only. The platform never modifies your resources. Fixing stays with you, and every hint points at your own remediation.
Metadata-only. Status codes, routes, hostnames, and bucket/object names. No request payloads, no query strings, no IP addresses.
Project-scoped. A project's feed can only ever show that project's own traffic. It is stored per project and isolated from every other tenant.

Glitch classes

Class	Meaning
`missing_bucket`	A storage read named a bucket that does not exist for the project. Recreate the bucket or fix the reference.
`missing_object`	An object inside an existing bucket was requested but not found. Check the file was uploaded.
`not_found`	Your app returned 404 (route or file gone).
`auth_error`	Your app returned 401/403 (may be intentional access control).
`rate_limited`	Your app returned 429 (its own limits, or Shield policy).
`server_error`	Your app returned 5xx. Check your logs.
`other_client_error`	Any other 4xx (400, 405, 410, 422).

Retention and Glitch Tracker Plus

The free feed keeps 7 days of history. Glitch Tracker Plus is an optional, explicitly opted-in service (org owner/admin, $1.40 per 1M events) that extends retention to 30 days and adds a weekly digest email. It never starts collecting before you enable it; the free feed is the part that was already on. Disabling Plus stops its collection within a minute, events collected before disabling are billed, and there are no refunds.

The weekly digest is fix-aware: it is only sent when there is something new or worsening, and it acknowledges resolved and improving issues rather than re-listing faults you are already fixing.

Console

Open your project, then Glitches in the sidebar. You can switch the window between 1, 7, and 30 days, expand any group to see sample paths, and read the recent-events list. The Plus opt-in card lives on the same page.

CLI

bash

# Aggregated summary for the project (default 7-day window)
scalix-cloud glitches get-glitch-summary
 
# Recent events, newest first
scalix-cloud glitches list-glitch-events

REST API

Both endpoints require a project-scoped token with the glitches:read scope.

Summary

bash

curl "https://api.scalix.world/v1/glitches?window_days=7" \
  -H "Authorization: Bearer $SCALIX_API_KEY" \
  -H "x-project-id: $SCALIX_PROJECT_ID"

Recent events

bash

curl "https://api.scalix.world/v1/glitches/events?limit=50" \
  -H "Authorization: Bearer $SCALIX_API_KEY" \
  -H "x-project-id: $SCALIX_PROJECT_ID"

MCP

The glitch feed is a REST and console surface. It is not currently exposed as an MCP tool; an MCP client cannot read it.