The whole product, on one page.
Errors, logs, performance and metrics, plus the parts that only matter once you are living in it: alerts that do not cry wolf, retention you control, and an API for everything the screens can do.
The four things it collects
Errors
Occurrences of the same bug become one issue with a title, a count, a level and a status. A resolved issue that happens again says so rather than opening a second one quietly.
Logs
Plain lines, searchable and filterable, with a live tail. Deliberately never grouped: folding two identical lines into one would misreport what happened.
Traces
Where a request or a job spent its time, with a waterfall for the traces worth keeping. Latency lives in histograms, so percentiles survive aggregation.
Metrics
Counters, gauges and histograms from any OpenTelemetry exporter, rolled up per hour. A cumulative counter is charted as change rather than as the running total it reports.
The parts that matter at three in the morning
Alerts with a reason
On a new issue, a regression, an occurrence threshold, a matching log line, or a measured number crossing a line. To email, a webhook, Slack or Teams.
Release health
A crash-free rate per release, which answers the question errors alone cannot: is this version worse than the last one.
Source maps
Upload the maps your build produced and a minified trace is shown as the code you actually wrote.
Browser noise control
An allowlist of your own script hosts, so an extension throwing on a visitors page does not become your issue, and what it refused is counted rather than hidden.
Absence is not zero
A metric alert never fires on a window with no readings, in either direction. An absent metric is not a metric below a threshold, and treating it as one turns every deploy gap into a page.
Issue ownership
Assign, comment, ignore, snooze until the next release. The activity trail says who did what.
The parts a procurement department asks about
A management API
Everything the dashboard does, from outside it, described by an OpenAPI document you can point a client generator at.
An MCP endpoint
So a coding assistant can read the issue it is being asked to fix, with a token whose scope you decide.
Roles
Owner, administrator, member and viewer, with invitations bound to the address they were sent to.
Your own storage
A bucket per organisation by default, or your own bucket with your own credentials.
Retention you set
Two windows per collector: how long an event row is kept, and how long its payload is. Your plan caps them; you can always ask for less.
Single sign-on
On Business and above, so leaving the company removes the access.
Fourteen days of Team, without a card.
Sign up with your work address, prove your domain with one DNS record, and paste a DSN. The first stack trace usually arrives before the coffee does.
No card. The trial ends by itself and drops to Free; nothing is charged unless you choose a plan.
