Free tool

A k6 load test, as a report anyone can sign off on.

Drop the JSON from k6 run --summary-export or handleSummary(). Get a pass/fail verdict from your thresholds, the numbers people actually ask about — p95, requests per second, error rate — every check by group, and the full timing table. Copy a Markdown summary for the pull request, or send the report to whoever decides if the release ships.

summary.json
4.9 KBhandleSummary JSON · 15 metrics
k6 load test report.html

2 of 4 thresholds failed · p95 613 ms · 0.71% failed requests · 98.93% checks passed

Publishes this report on Comma: one URL that outlives the CI artifact, restricted to named people or your company’s email domain if you like, comments pinned to the threshold that failed, and the next run kept as a revision you can diff. Free, no card.

How the summary is read

  • The verdict comes from the thresholds on each metric — the same rule that makes k6 exit with code 99. Failed thresholds are listed first, with the measured value next to the limit.
  • Headline numbers are http_reqs (count and rate), http_req_duration p95, http_req_failed, the overall checks pass rate, iterations, max VUs and the run length from state.testRunDurationMs.
  • Checks are collected from root_group and every nested group, failing checks first, labelled with their group path.
  • Timings show every trend metric and submetric with whichever stats you configured in summaryTrendStats — p(99), p(99.9) included.
  • Older exports from --summary-export carry no metric types and mark a threshold true when it failed; both are handled, so a passing run isn’t shown as red.

Getting the JSON

FromHow
CLI flagk6 run --summary-export=summary.json script.js
handleSummaryexport function handleSummary(data) { return { 'summary.json': JSON.stringify(data) }; }
More percentilesk6 run --summary-trend-stats='avg,med,p(95),p(99)' …
GitHub Actionsgrafana/run-k6-action, then upload summary.json with actions/upload-artifact
Dockerdocker run --rm -v $PWD:/w -w /w grafana/k6 run --summary-export=…

If handleSummary is defined, k6 stops printing its own terminal summary unless you also return stdout — add stdout: textSummary(data) from the k6-summary jslib to keep it.

When the load test has to leave the terminal

A load test is usually run by one engineer and judged by several people: the PM who wants to know if launch day traffic is safe, the SRE who sets the capacity plan, the client who was promised a p95. A pasted terminal summary loses its alignment in Slack and a CI artifact expires. Published to Comma, the report is one link with comments on the threshold that failed, and an agent can publish it from the pipeline after every run, as a new revision of the same report — so this week’s p95 sits next to last week’s.

Questions

How do I turn a k6 summary into an HTML report?
Run k6 run --summary-export=summary.json script.js, or add a handleSummary(data) function that returns { 'summary.json': JSON.stringify(data) }, then drop that file here. You get a pass/fail verdict from your thresholds, the headline numbers (requests, requests per second, p95, failed requests, checks), every threshold with the value it was tested against, every check by group, and the full timing table. Download it as one HTML file, copy a Markdown summary, or publish it as a link.
Which k6 files does it read?
The JSON handleSummary() receives (k6 v0.30 and later, including k6 v1), and the older --summary-export file, whose metrics have no type field and whose thresholds are true when they failed — both are read correctly. Submetrics such as http_req_duration{name:POST /api/checkout} get their own rows. The raw --out json stream is one line per sample rather than a summary, so the tool tells you which flag to use instead.
Is this the same as k6-reporter or the k6 web dashboard?
It shows the same summary data without changing your script: k6-reporter needs an import and a handleSummary hook, and the web dashboard (K6_WEB_DASHBOARD=true) only lives while the test runs unless you export its report. Here you drop the JSON you already have — from last night's CI run, say — and nothing is installed.
Why did k6 exit with code 99?
Exit code 99 means at least one threshold failed. The report puts failed thresholds first with the actual number next to the limit — for example p(95)<500 against a measured 612.8 ms — so you can see whether it missed by a hair or by a mile. Checks failing does not change the exit code on its own; add a threshold on the checks rate (rate>0.99) if it should.
Why is http_req_failed shown as a percentage of “true” values?
http_req_failed is a rate metric: each request adds true if it failed and false if it didn't, so its rate is the share of failed requests and passes counts failures. That inversion trips people up in k6's own terminal output; the report labels it as failed requests.
Can I post the results on a pull request?
Yes. “Copy Markdown” gives a GitHub-flavored verdict line, the headline numbers, a threshold table with ✅ and ❌, and the failing checks — it pastes into a PR comment or appends to $GITHUB_STEP_SUMMARY. Link the published report for the timing table.
Is my summary uploaded?
No. The JSON is parsed and the report is built in your browser. It only leaves your machine if you click “Get a shareable link” and publish it to Comma — and then you choose who can open it.

Related: Lighthouse viewer · JUnit XML viewer · Postgres EXPLAIN visualizer · JSON to chart · all free tools