Core Web Vitals issues API with what to fix first, from your real users
Every issue CoreDash finds in your real-user data, with its fix, plus release checks and budgets, as JSON for your scripts and your AI.
A lazy-loaded LCP image on your product pages, a slow script host behind INP on mobile, a budget that went over: CoreDash finds them every day. After each release it checks whether LCP, INP and CLS got worse, and shows p75 before and after, so you see whether it helped. Your CI and your AI read the answer instead of working it out from percentiles.
Endpoints for issues, releases and budgets
Send each call to https://app.coredash.app followed by the path in the table. A project key reads its own site and records releases from your pipeline. An account key reads every site in your account and does the writes: issue actions, changes and budgets.
Fix first
The 5 open issues with the most impact or, when none is open, the 5 pages furthest over the Good line.
The fix with every issue
Every issue says why it matters, how to fix it and the rule that found it. Mark it fixed and CoreDash checks it again after 7 days.
Release check
Your pipeline posts the release tag with a project key and reads pass or worse, with the findings per device.
Budgets
Your own line per metric and segment. A line that goes over opens an issue.
| Endpoint | MCP tool | Access |
|---|---|---|
GET/api/ | None | Read |
GET/api/ | list_changes | Read |
GET/api/ | get_change | Read |
POST/api/ | None | Account key |
PATCH/api/ | None | Account key |
DELETE/api/ | None | Account key |
POST/api/ | None | Project key |
GET/api/ | get_release_check | Read |
GET/api/ | list_budgets | Read |
POST/api/ | None | Account key |
PATCH/api/ | None | Account key |
DELETE/api/ | None | Account key |
GET/api/ | list_issues | Read |
GET/api/ | get_issue | Read |
POST/api/ | None | Account key |
GET/api/ | get_fix_first | Read |
The demo site's slowest pages, from one call
The demo project had no open issue when this answer was fetched, so fix first lists its slowest pages instead: per URL the Core Web Vital furthest over its Good line, at p75.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/fix-first"AnswerJSON
{
"issues": [],
"total": 0,
"pages": [
{
"url": "https://www.corewebvitals.io/state-of-cwv/explorer/technologies/cdn/by-fcp/all",
"metric": "CLS",
"value": 1,
"views": 75
}
],
"percentile": "p75"
}What an issue holds, with the rule that found it
The issue list returns the open issues: Open, Came back and Fix applied. Ask for closed or ignored ones with status, and for one device with d. Each issue names its check, what it was found on (a page group, an element, a script host or the whole site), the device and its status. It carries the current value with its page views and its impact, and one value per day it was found, for up to 90 days. Next to the list come the counts per status and the time of the last run.
Open one issue, and its help holds name, why, fix and rule. The rule text reads its thresholds from the same config the check runs on. Then come the actions the issue allows right now, and its metric as a summary and a series over the last 7 days, with the change markers in that period. A Lighthouse budget issue has no real-user chart, so its summary and series are null and it has no markers. Up to 5 page views that CoreDash sampled in full (the deep scan) and that match the issue come with it as evidence, each with its waterfall.
"What should I fix first on mobile?"
Connect CoreDash at https://app.coredash.app/ and ask your AI. It calls get_fix_first for mobile and gets the open issues with the most impact. Say the top one is a lazy-loaded LCP image on your product pages. get_issue gives it the mechanism: the browser fetches a lazy image only after layout, so the LCP request starts late and Load Delay grows. The fix comes with it: drop loading="lazy" from the image that becomes the LCP element, lazy-load from the second or third image down, and add fetchpriority="high". Your AI makes that change in the template behind those pages.
Open issues in your ticket tracker every Monday
A weekly script reads the open issues and opens a ticket for every issue that has none yet, with the issue id in the ticket. Once the change is live, the script sends fix_applied to the issue with an account key, which also puts a Fix applied marker on the charts. CoreDash then waits 7 days from that call, so its window holds only page views from after it. A finding that is gone makes the issue Fixed. One that is still there sends it back to Open. Next Monday your script sees Open where it left Fix applied, and reopens the ticket.
So send it on the day the change goes live. A call before the deploy leaves page views of the old code in the window, and a fix that works can go back to Open. Sent late, it delays the verdict by just as long.
An issue the release check opened follows that check instead. The check of your next release closes it when the metric is lower again, and sends a Fix applied issue back to Open when that release is worse too.
Did today's release make INP worse?
Your deploy job records the release with its tag and a project key. One release per tag, so a retry updates it, and the answer links to its release check. A second job then asks the release check for that tag every 15 minutes until final is true. On a quiet site that takes up to 24 hours, so give it a job of its own after the deploy. A deploy step that waits for it holds your pipeline that long. The status reads collecting, pass, worse or insufficient, and insufficient means no device got enough page views. Fail that job on worse.
When your page views carry the release tag, CoreDash compares them with the page views of the release before. Without the tag it compares the hours after the release with as many hours before it. The check is done once mobile and desktop each have 500 page views on both sides, and at the latest 24 hours after the release or at your next release. A device with only a small share of your page views is left out. When the result is worse, CoreDash opens an issue for that device, on the whole site or on a busy page group.
Every release with p75 before and after
The list of changes holds your releases, notes and applied fixes, newest first. The newest 40 carry LCP, INP and CLS at p75 before and after, so you see whether last week's release helped. The markers call gives the same changes for any period, to draw on your own charts.
A metric that slides without any release is the daily run's job: it compares the last 7 days with the 28 days before. A sudden step usually points at a deploy, a slow slide at third-party scripts piling up.
A budget that goes over becomes an issue
The budgets call lists your budgets with the state of every line: over budget, at the budget, within it, or too few page views to judge. An account key adds a budget, with a line per metric, for a device and any dimension filters. Every line is judged over the last 7 days and needs 200 page views.
Each line holds when p75 of its segment stays under it. One that sits right on its budget crosses it by chance on some days, so it counts as over only when p75 is above the line by more than chance explains for its page views, and by more than a small margin per metric. Then the daily run opens an issue. That issue stays open until the line is clearly under the budget again, so it does not open and close every other day.
A Lighthouse budget that goes over opens an issue too, with a name that ends in (lab), so your CI and your report read lab and field problems from one list. The monitors and runs are on the Lighthouse API.
Poor issues first in the client report
Your monthly report script reads the open issues of each client site on mobile and puts the Poor ones at the top: a slow script host behind INP on a webshop's checkout comes before a background image on a blog. Each line carries the page or element and its impact. The counts per status give the progress line under it, open against closed and ignored, so the client sees last month's fixes counted too.
Issues API reference
The parameters and answer fields of every endpoint and MCP tool in this topic, generated from the code.
GET/api/v1/projects/{projectId}/markers
Lists the release and note markers that the charts show for a period.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
datequery, stringThe period: -6h, today, -1d, -7d, -28d or -56d.
fromquery, stringThe start of the period as an ISO date and time, with to.
toquery, stringThe end of the period as an ISO date and time.
startdatequery, stringThe first day, as YYYY-MM-DD in UTC, with enddate.
enddatequery, stringThe last day, included.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/markers?date=-7d"AnswerJSON
{"start": 1794149436444, "end": 1794754236444, "markers": []}GET/api/v1/projects/{projectId}/changes
Releases and notes, newest first, each with p75 of LCP, INP and CLS before and after it.
MCP tool: list_changes
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
dquery, stringDefault: all.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/changes"AnswerJSON
{
"changes": [
{
"id": "r_6a6f09721200168bdd7f62c8",
"kind": "release",
"at": 1785661810000,
"title": "v2.1.0-prod-202608020910",
"tag": "v2.1.0-prod-202608020910",
"notes": "prod latestRelease for createjs",
"source": "ci",
"meta": {},
"editable": false
}
],
"windows": {},
"deltas": {}
}GET/api/v1/projects/{projectId}/changes/{changeId}
One release or note: its numbers before and after, the change markers around it and, for a release, its release check with the issues it links.
MCP tool: get_change
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
changeIdpath, stringrequiredThe id of the release or note, from the list of changes.
dquery, stringDefault: all.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/changes/r_6a6f09721200168bdd7f62c8"AnswerJSON
{
"change": {
"id": "r_6a6f09721200168bdd7f62c8",
"kind": "release",
"at": 1785661810000,
"title": "v2.1.0-prod-202608020910",
"tag": "v2.1.0-prod-202608020910",
"notes": "prod latestRelease for createjs",
"source": "ci",
"meta": {},
"editable": false
},
"numbers": null,
"check": {
"status": "insufficient",
"final": true,
"basis": "time",
"tag": "v2.1.0-prod-202608020910",
"against": null,
"at": 1785661810000,
"after": [1785661810000, 1785748210000],
"before": null,
"checkedAt": 1794754236858,
"devices": [],
"findings": []
},
"issues": {},
"markers": []
}POST/api/v1/projects/{projectId}/changes
Adds a release or a note.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
kindbody, stringrequiredrelease or note.
titlebody, stringThe title of a note. A note needs one.
tagbody, stringThe tag of a release: letters, digits, dots, underscores, slashes and hyphens. A release needs one.
atbody, stringWhen it happened, as an ISO date. Default: now.
notesbody, stringFree text.
PATCH/api/v1/projects/{projectId}/changes/{changeId}
Changes a release or a note that was added in CoreDash or by CI. Send the whole change again: a field left out is cleared, and without at it moves to now.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
changeIdpath, stringrequiredThe id of the release or note, from the list of changes.
titlebody, stringThe title of a note. A note needs one.
tagbody, stringThe tag of a release: letters, digits, dots, underscores, slashes and hyphens. A release needs one.
atbody, stringWhen it happened, as an ISO date. Default: now.
notesbody, stringFree text.
DELETE/api/v1/projects/{projectId}/changes/{changeId}
Deletes a release or a note.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
changeIdpath, stringrequiredThe id of the release or note, from the list of changes.
POST/api/v1/projects/{projectId}/releases
Records a release from your build pipeline with a project key. One release per tag, so a retry updates it. The answer links to the release check.
Use a project key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
tagbody, stringrequiredThe release tag: letters, digits, dots, underscores, slashes and hyphens.
deployedAtbody, stringWhen it was deployed, as an ISO date. Default: now.
shabody, stringThe commit.
branchbody, stringThe branch.
repobody, stringThe repository.
prNumberbody, integerThe pull request number.
actorbody, stringWho deployed it.
runUrlbody, stringThe address of the pipeline run.
notesbody, stringFree text.
GET/api/v1/projects/{projectId}/releases/check
The release check of one release, found by its tag: status collecting, pass, worse or insufficient, the findings, and final once the result no longer changes.
MCP tool: get_release_check
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
tagquery, stringrequiredThe release tag, for example v2.4.0.
GET/api/v1/projects/{projectId}/budgets
The performance budgets with the status of every line: over budget, at the budget, within it, or too few page views to judge.
MCP tool: list_budgets
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
dquery, stringDefault: all.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/budgets"AnswerJSON
{
"budgets": [
{
"budget": {
"id": "l_6abe7dfcb303d3cd97bfdc86",
"name": null,
"device": "mobile",
"filters": {"u": "https://www.corewebvitals.io/"},
"lines": {"LCP": 1565, "INP": 119, "CLS": 0.088, "FCP": 958, "TTFB": 541},
"legacy": true
},
"lines": [
{
"metric": "LCP",
"budget": 1565,
"p75": 708,
"views": 24,
"within": 0.9583333333333334,
"state": "few"
}
]
}
]
}POST/api/v1/projects/{projectId}/budgets
Adds a budget with one or more lines.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
namebody, stringA name for the budget.
devicebody, stringmobile, desktop or all. Default: all.
filtersbody, objectDimension keys with a value, to limit the budget to those page views.
linesbody, objectrequiredThe limit per metric, keyed by LCP, INP, CLS, FCP and TTFB.
PATCH/api/v1/projects/{projectId}/budgets/{budgetId}
Changes a budget. Send the whole budget again.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
budgetIdpath, stringrequiredThe id of the budget, from the list of budgets.
namebody, stringA name for the budget.
devicebody, stringmobile, desktop or all. Default: all.
filtersbody, objectDimension keys with a value, to limit the budget to those page views.
linesbody, objectrequiredThe limit per metric, keyed by LCP, INP, CLS, FCP and TTFB.
DELETE/api/v1/projects/{projectId}/budgets/{budgetId}
Deletes a budget.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
budgetIdpath, stringrequiredThe id of the budget, from the list of budgets.
GET/api/v1/projects/{projectId}/issues
The issues CoreDash found, as the Issues page lists them: per issue the check, the subject (a page, an element), the device, the status and the current value with page views and impact. Also the count per status and the last run.
MCP tool: list_issues
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
statusquery, stringopen, closed or ignored. Default: open.
dquery, stringDefault: all.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/issues"AnswerJSON
{
"issues": [],
"counts": {"open": 0, "closed": 0, "ignored": 0},
"run": {"lastRunAt": 1794613035175, "percentile": "p75", "failed": []}
}GET/api/v1/projects/{projectId}/issues/{issueId}
One issue: its numbers and history, the help text (what the check found and how to fix it), the chart of its metric, the change markers around it and the sampled page views that show it.
MCP tool: get_issue
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
issueIdpath, stringrequiredThe id of the issue, from the list of issues.
There is no example answer for this call.
POST/api/v1/projects/{projectId}/issues/{issueId}
Marks an issue as fixed, ignored or open again.
Use an account key for this call.
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
issueIdpath, stringrequiredThe id of the issue, from the list of issues.
actionbody, stringrequiredfix_applied, ignore or reopen.
GET/api/v1/projects/{projectId}/fix-first
What to fix first, as the Overview shows it: the open issues with the most impact, or, when no issue is open, the slowest pages in the period.
MCP tool: get_fix_first
Parameters
projectIdpath, stringrequiredThe id of the site. The list of sites gives it.
dquery, stringDefault: all.
datequery, stringPeriod. Default: -28d.
percentilequery, stringPercentile of each point. Default: p75.
curl -H "Authorization: Bearer cdk_your_key" \
"https://app.coredash.app/api/v1/projects/your_project_id/fix-first"AnswerJSON
{
"issues": [],
"total": 0,
"pages": [
{
"url": "https://www.corewebvitals.io/state-of-cwv/explorer/technologies/cdn/by-fcp/all",
"metric": "CLS",
"value": 1,
"views": 75
}
],
"percentile": "p75"
}Every client site with one account key
An account key reads the issues of every site in your account and marks them, so one script runs the ticket sync or the report for all your clients. See the agency API and agency pricing.
Keys, the MCP server and the RUM numbers
Frequently asked questions
How often does CoreDash run the checks?
Once a day, for every site with more than 500 page views in the last 7 days. Each check reads those 7 days, and the checks for a metric that got worse compare them with the 28 days before. The release check runs apart from that, after each release.
Which problems does CoreDash check for?
Known causes per metric, read from the attribution in your real-user data. LCP checks look at how the LCP image is loaded and which sub-phase takes most of the time. INP checks find the element and the script host behind slow interactions, and CLS checks the element that shifts. For TTFB it is the server, the connection, and the URLs or visitors that wait longer. Then low use of the back/forward cache and the gap between TTFB and FCP. On top of those come the checks that compare: a metric that got worse, a release that made one worse, and a budget that went over, in real-user data or in Lighthouse. Every issue names its check, and its help holds the exact rule.
Which key does my CI need?
A project key. It records the release and reads the release check. Marking an issue fixed, ignored or open again, and writing changes or budgets, take an account key. No MCP tool marks an issue, so an AI does that through the API with an account key.
What happens when I mark an issue fixed?
CoreDash checks it again at the first daily run 7 days after your call, when the window holds only page views from after it. Until then the issue shows Fix applied. If the finding is gone, it becomes Fixed; if it is still there, it goes back to Open as a fix that failed. An issue from a release check waits for your next release instead: that release's check closes it when the metric is lower again. A fixed issue that turns up again later becomes Came back.
Does an issue close by itself?
Yes, most do. After 3 daily runs in a row without the finding, an open issue becomes Auto closed. If a later run finds it again, it becomes Came back and returns to the open list. Two kinds work differently. An issue from a release check closes when a later release makes the metric lower again. A budget issue stays open until the line is clearly under the budget again.
Why is there no issue for my slow page?
Usually because no check matched it, or because bigger findings filled the list: a check keeps its 15 biggest findings per device. Most checks look at a page group as a whole, so one slow URL can sit in a group that passes. The site also needs more than 500 page views in the last 7 days to be checked at all. For a slow page without an issue, read fix first: with no open issue it lists the 5 pages furthest over the Good line on LCP, INP or CLS, among the 100 URLs with the most page views, each with at least 20 page views.
Can I get new issues in Slack or by email?
Yes. Email, Slack and webhooks can each get new issues (optionally only the Poor ones), releases that made one of LCP, INP and CLS worse, budgets that go over, and issues that come back or get fixed. You pick the channels and the kinds in the CoreDash app.
Does the issues API cost extra?
No. The API and the MCP server come with every paid plan and every running trial. A key makes 60 calls a minute by default, up to 960 on the largest plans.