Skip to main content

Send Slack alerts when Playwright tests fail

The usual version of this is a conditional step at the end of a CI job that shells out to curl with a hand-built JSON payload, duplicated across every workflow file that needs it. This does the same job as an edge on the canvas.

What you need​

Connect Slack​

The Slack integration authenticates either through OAuth or an incoming webhook. OAuth is the better default: it lets Playrunner list your channels, so the node configuration becomes a channel picker rather than a pasted URL.

Connect it once from settings. The connection is reused by every workflow.

Build the workflow​

The graph is three nodes:

Environment ──sequential──▶ Playwright ──failure──▶ Slack
  1. Environment node — the variables your suite needs, defined once and injected into the run.
  2. Playwright node — the repository, branch, and Playwright version to run.
  3. Slack node — the channel and the message.

The important part is the edge type between Playwright and Slack. Draw an On Failure connection. The Slack node then runs only when the Playwright node fails, and is skipped when it passes. No condition inside the node, and nothing to keep in sync.

Route passing and failing runs differently​

Most teams want failures loud and passes quiet. Draw two edges out of the same Playwright node:

                   ┌──failure──▶ Slack  (#qa-alerts)
Playwright ────────┤
                   └──success──▶ Slack  (#qa-runs)

Both nodes exist on the canvas, each configured with its own channel and message. Exactly one of them runs per execution, because On Success and On Failure are mutually exclusive.

If you want a message on every run regardless of outcome, use an Independent connection instead — it runs after the parent finishes either way.

What to put in the message​

Keep it short enough to read in a notification and useful enough to act on:

  • Which workflow ran, and on which branch.
  • Whether it passed or failed.
  • A link back to the run, so the Playwright report, screenshots, and traces are one click away.

Artefacts stay attached to the run rather than being uploaded into Slack, so the channel does not become a dumping ground for zip files.

Common variations​

  • Escalate only repeated failures — put the Slack node behind a second workflow triggered on a schedule rather than on every run.
  • Summarise before alerting — insert an AI triage node between Playwright and Slack, so the message says what broke rather than that something broke.
  • Open a ticket at the same time — draw a second failure edge to a Jira node. Both run on the same failure.

Next steps​