inline

Runtime errors, rooted in the code

inline collects the crashes your users hit, groups them into one issue per distinct bug, and traces each one back to the code that shipped it. Link a project to a repository and the root-cause analysis names the change, using the same read-only tools the reviewer uses.

One issue
per bug

An SDK sends errors in. A crash that happened four thousand times is one issue with a count on it, showing how many events and how many people it affected and when it was last seen. Upload source maps and stack traces come back symbolicated.

Issuesacme-web · production
Unresolved

12

Events 24h

8,104

Users

431

TypeError: cannot read 'tenantId' of null
1,204 events · 88 users · P1
root cause
AbortError: signal is aborted without reason
612 events · 140 users · P2
regressed
PrismaClientKnownRequestError P2002
88 events · 12 users · P2
triaged
ChunkLoadError: loading chunk 42 failed
41 events · 39 users · P4
noise

How it works

Priority from triage,
not from volume

Each issue gets a priority from P1, fix now, to P4, noise. Triage reads the error, which is why a rare crash in checkout can outrank a common one on a settings screen. Regressed is the number to watch: an issue you had closed has started happening again.

The change
that shipped it

Open an issue and inline gives you its analysis of what went wrong in the code: it points at the release commit, finds the offending line and, where it can, writes the patch. Issues with a proposed fix are marked and counted, so a quick win does not sit in the backlog.

Defaults you
set once

Turn triage and root-cause analysis on or off for the workspace, set a minimum level for what is kept at all, and choose how long events are retained. An event volume chart over 24 hours, 7 days or 30 days tells a new bug from one that has always been there.

Included

SDK ingestevents sent with the project's public key
Source mapsuploaded with the project's write key
P1 to P4priority from triage, not raw counts
Regressionsclosed issues that came back, counted on their own
Suggested fixesa patch written where the cause is clear
Retentionyou choose how long events are kept

Put it on one repository

Connect GitHub or GitLab, enable a repo, and the next pull request opened gets a review.