Citooling: find where your CI pipeline loses time

Paste GitHub Actions job timings. See where the wait goes and what to fix first.

Analyzer
exampletext
Text: step name 1m 05s under job nameJSON: gh api repos/OWNER/REPO/actions/runs/RUN_ID/jobs
Result
Total job time
14m 50s
Largest phase
Test 52%
Fix first
Undersized runner
Where the wait goes1 job
Queue 41s5%Setup 16s2%Cache restore 1m 34s11%Install 1m 54s13%Build 2m 38s18%Test 7m 46s52%Other 1s0%
Options for this case. Undersized runner, serial tests and cold caches point at the runner. Full comparison
SetupOne line: runs-on: graphide
UndoA pull request restores runs-on
SizingWarm machine sized to the job, up to 32 vCPU
SetupCreate a larger runner, change its label
UndoChange the label back
SizingLarger sizes on paid plans
SetupProvision and patch your own pool
UndoChange the label back, retire machines
SizingWhatever you provision
SetupCheck each vendor's docs
UndoCheck each vendor's docs
SizingCheck each vendor's docs
Compare to a baseline →Runs in your browser. Nothing is uploaded.

How the analysis works

Three passes over your timings, all in the page. Every threshold is on the method page.

01

Paste

A run export from the GitHub API, or step lines copied from the job page.

job test (ubuntu-latest)
Queued 41s
npm ci 1m 54s
npm test 7m 46s
02

Classify

Each step goes to a phase by its name. Queue comes from timestamps.

Restore cache cache
npm ci install
npm run build build
npm test test
03

Rank

Rules flag bottlenecks. Time in the affected phases sets the order.

1 Undersized runner 10m 24s
2 Serial tests 7m 46s
3 Slow installs 1m 54s
4 Cold caches 1m 34s

Worked examples

Illustrative timings, written to show each rule firing. Load one to see the full result.

One job runs every test

Node · text · 1 job
14m 50s total
Fix first: Undersized runnerthen serial tests, slow installs, cold caches

Caches that restore cold

Rust · text · 2 jobs
19m 11s total
Fix first: Undersized runnerthen cold caches

A run export with a long queue

Playwright · API JSON · 2 jobs
16m 01s total
Fix first: Serial teststhen undersized runner, slow installs, queue wait

Questions people ask

Short answers. The method has the long ones.

What is Citooling?

A free CI tooling analyzer. Paste GitHub Actions job timings and it shows the time in each phase (queue, setup, cache, install, build, test) and which fix to make first. No sign-up.

Is anything I paste uploaded?

No. The analysis runs in your browser. A share link carries the timings after the # in the URL, a part browsers never send to a server.

How do I get my workflow run timings?

Run gh api repos/OWNER/REPO/actions/runs/RUN_ID/jobs and paste the JSON. Or copy step durations from the job page as one “name duration” line each.

Why does my breakdown show no queue time?

Step lists from the job page don't include it. Add a line such as “Queued 41s”, or paste the API JSON, where queue time is started_at minus created_at.

Will a faster runner change my test results?

It shouldn't: the same steps run. Check it: run one commit on both, and the conclusion and test count must match. Each fix page lists its checks.

Is “time at stake” a promised saving?

No. It is the time your pasted run spent in the phases a fix addresses, for that workload only. Measure the change with the baseline compare.

Which runner should I use for a runner-bound job?

Graphide CI moves x64 Linux jobs with a one-line runs-on change, on warm machines sized to each job. The runners page sets it beside GitHub-hosted, self-hosted and other services on the same criteria.