
upkeep-mcp
io.github.tiagocalado600/upkeep-mcp
Website maintenance checks: domains, SSL, uptime, technical SEO and accessibility
Documentation
upkeep-mcp
An MCP server for the recurring checks behind ongoing website maintenance: domains, SSL certificates, uptime, technical SEO and accessibility — all from publicly available information.
Built for people who look after a portfolio of client sites on a retainer, not just a single domain. The goal is to answer one question quickly: what needs attention this week?
Status
Early development, built in public phase by phase. Every planned tool is
implemented, accessibility_audit included, and useful today.
| Tool | Purpose | Status |
|---|---|---|
domain_check | Registration expiry, registrar, nameserver agreement, DNS records, DNSSEC | Available |
ssl_check | Certificate expiry, issuer, chain validity, SAN coverage, TLS version | Available |
uptime_check | HTTP status, response time, redirect chain, HTTPS upgrade, security headers | Available |
health | Server name, version, Node.js version, uptime | Available |
seo_audit | Title, meta, headings, canonical, robots.txt, sitemap, broken links | Available |
site_crawl | Duplicate titles, broken links and stray noindex across a whole site | Available |
portfolio_report | All of the above across a portfolio, sorted by urgency | Available |
accessibility_audit | WCAG violations via axe-core, in a real browser | Available |
It also exposes the portfolio://sites resource (the site list, for a client to
read without spending a tool call) and the quarterly_report prompt (turns a
portfolio run into the report a client actually reads).
Published on npm, so it installs with one command — see Installation. A public instance is also running, for anyone who would rather point a client at a URL than run anything — see The hosted instance.
What it looks like
> Is example.com about to expire?
example.com expires 2027-08-13 (268 days).
Registrar: RESERVED-Internet Assigned Numbers Authority.
Nameservers: elliott.ns.cloudflare.com, hera.ns.cloudflare.com (2 of 2 answered,
all on serial 2420077714).
Resolves: apex yes, www yes. DNSSEC: signed and validating.
Email: SPF -all, DMARC p=reject.
Needs attention:
- [info] DMARC publishes no "rua" address, so no reports arrive to show whether
it is working.
> Check the certificate on expired.badssl.com
expired.badssl.com:443 certificate expires 2015-04-12 (-4238 days).
Issued by COMODO RSA Domain Validation Secure Server CA.
Chain does not verify (CERT_HAS_EXPIRED). Negotiated TLSv1.2.
Host matched via *.badssl.com.
Revocation not established: http://ocsp.comodoca.com could not answer: the
responder is not authorised to answer for this certificate.
Needs attention:
- [critical] The certificate expired 4238 days ago.
- [critical] The certificate chain does not verify: CERT_HAS_EXPIRED.
- [unknown] Revocation could not be checked: http://ocsp.comodoca.com could not
answer: the responder is not authorised to answer for this certificate.
> Check the certificate on revoked-ecc-dv.ssl.com
revoked-ecc-dv.ssl.com:443 certificate expires 2026-12-24 (36 days).
Issued by SSL.com TLS Issuing ECC CA R2.
Chain verifies. Negotiated TLSv1.3.
Host matched via revoked-ecc-dv.ssl.com.
Revoked on 2026-06-09.
Needs attention:
- [critical] The certificate was revoked on 2026-06-09; browsers that check
revocation will refuse the site.
- [info] This certificate is valid for 198 days, so it is probably renewed by
hand. New public certificates are limited to 200 days, 100 from 2027-03-15 and
47 from 2029-03-15, so that renewal is worth automating.
Note what the revoked one does: the chain verifies. Node performs no revocation lookup of its own, so that certificate completes a handshake and reports as trusted. Only asking its issuer finds the problem.
Every tool also returns structured data alongside the text, so results can be
sorted, filtered and fed into a report. Full input and output for each tool is in
examples/, and a whole portfolio session — the weekly triage, a
drill-down, and what the comparison against the previous run will and will not
claim — is in examples/conversation.md.
The tools
domain_check
Input: domain — a bare domain, a full URL, or an internationalised name. A
subdomain is reduced to its registrable domain, since that is what a registration
belongs to. A bare public suffix such as co.uk is refused as invalid input —
here, in ssl_check and in a portfolio's domain — since nobody owns one.
Returns the expiry date and days remaining, the registrar and its IANA ID,
registry statuses, A/AAAA/NS/MX/TXT/CAA records, whether the apex and www
resolve, whether the delegation is signed with DNSSEC, and what the domain's SPF
and DMARC records say about who may send email as it.
Email authentication is read from the domain's own DNS — the SPF record at the
apex, the DMARC record at _dmarc. A record that is absent is reported as
information, because it is a standing improvement rather than something that
broke this week. A record that is present and wrong is a warning, because it
fails right now: two SPF records make receivers skip SPF entirely, and +all
authorises the whole internet to send as the domain. A record the resolver
could not read — a TXT answer too large for UDP, retried over TCP to a home
router that refuses TCP, is the usual cause — is never reported as absent: it is
unknown, with the resolver's reason, and the same goes for any other record set
that went unanswered while the rest of the lookup succeeded.
DKIM is deliberately not reported. A DKIM key lives at <selector>._domainkey,
and a selector cannot be discovered — only guessed, one DNS query per guess.
That is subdomain enumeration, which this project does not do, so a domain with
no DKIM and one whose selector was not guessed are left indistinguishable rather
than the second being reported as the first.
Each of the domain's own nameservers is then asked about the zone directly,
over TCP port 53 with recursion off, which is the one question a recursive
resolver cannot answer: it replies with whatever one server told it and does not
say which. That finds a server left in the delegation after a migration — it
answers REFUSED, or its own hostname stopped resolving, and every resolver
query landing there is slow or fails — and it finds nameservers holding
different versions of the zone, which is the "it works for me but not for my
colleague" outage.
What is a fault and what is only unestablished are graded apart. Resolvers ask
over UDP first and this server can only use TCP, so a nameserver that refuses
TCP is not a broken one: sapo.pt's four all refuse it and the domain resolves
perfectly. Different serials are info too — github.com runs two providers
that do not transfer between them, so four of its nameservers report
1656468023 and four report 1, and nothing is wrong. A hostname that does not
resolve, or an answer without authority for the zone, is broken for everybody
and is a warning. Pass checkNameservers: false to skip the whole thing —
portfolio_report does, because a portfolio would pay this per site and a
deployment that cannot open port 53 would grade every site at once as
unestablished.
docs/adr/0020 records why
it speaks DNS by hand and what it deliberately does not check.
ssl_check
Input: domain, optional port (443 by default).
Returns expiry and days remaining, issuer, whether the chain verifies and why not when it does not, which hostnames the certificate covers and via which SAN entry, the negotiated TLS version and cipher, and whether the certificate has been revoked. Expired, self-signed and untrusted certificates are inspected and reported rather than refused — those are the ones worth finding.
A certificate that covers the bare domain but not www (or the reverse) is not
by itself a fault: many sites, cloudflare.com among them, serve www from a
certificate of its own. So when the host asked about is one of the two and the
other resolves but is not covered, one more handshake is made with the other
name, and only what it actually serves decides the finding — www_not_covered
when that is not valid for it, www_https_failed when no secure connection can
be made there at all, www_coverage_unjudged when it could not be established
(and the apex_ equivalents from the other side).
Revocation is checked over OCSP. The response a server staples to the handshake
is preferred, because it costs no request at all; when there is none, the
responder named in the certificate is asked directly. An answer is only believed
once its signature verifies against the issuing authority, only once its
CertID is shown to be about the certificate that was actually served — a server
serving a revoked certificate alongside a valid response for a different one is
otherwise the easy way to fake a clean result — and only while it is current,
allowing five minutes of clock skew either side. A staple that fails any of these
is not the end of it: the responder is asked anyway, so a server cannot suppress
its own revoked by stapling something unusable.
Many healthy certificates cannot be checked at all: since 2025 the two largest
issuers, Let's Encrypt and Google Trust Services, publish no OCSP responder and
distribute revocation by CRL. That is reported as an unavailableReason and
produces no finding, because it is a decision of the certificate authority
and nothing the site owner can act on. A responder that was asked and would not
answer is different, and gets one unknown.
uptime_check
Input: url — a full http or https URL, or a bare domain, which is tried over
HTTPS. Any other scheme, such as ftp:, is refused as invalid input.
Returns the status code, response time, every hop of the redirect chain, whether plain HTTP is upgraded to HTTPS, the HSTS policy, and the security headers worth reporting on. A site whose certificate browsers refuse — expired, issued for another host, self-signed — is reported as down, since no visitor gets past the warning.
seo_audit
Input: url — the page to audit, plus optional checkLinks (true by default)
and maxLinks (25 by default).
Returns title and meta description with their lengths, the heading structure,
canonical, lang, viewport, Open Graph, hreflang alternates, the images with
no alt attribute, the state of robots.txt and the sitemap, and which
internal links are broken.
Only a sitemap URL that answers 404 or 410 is reported as no sitemap, an
info gap. One that answers 403 or a server error may well exist and is
unreadable to search engines too, so it is a warning that the sitemap could
not be read. One that does not answer at all — a timeout, a failed connection —
says nothing about the sitemap, so it is graded unknown: the check failed,
not the site.
A gzipped sitemap is unpacked before it is read. Whether it is gzipped is decided
by the first two bytes rather than by the file name or the content type, because
plenty of files called .xml.gz are not, and plenty that are get labelled
text/xml.
The sitemap is then checked against the rules of the protocol, because a file
that answers 200 and parses is not the same as a file that works: a root
element with no sitemaps.org namespace is dropped whole, an entry on another
host is discarded, and one unescaped & makes the document ill-formed XML and
costs every entry after it. Each broken rule is reported once with the number of
entries that break it and one example, so a mistake repeated across fifty
thousand URLs reads as one thing to fix. A rule that only costs a hint — a
<lastmod> that is not a W3C Datetime, a <changefreq> outside its seven
values — is graded info; one that costs the entry or the file is a warning.
robots.txt is read before anything else is requested and is obeyed — for
the page itself and for every internal link. A page this crawler is not allowed
to read is reported as such and is never fetched, and an unreadable robots.txt
is treated as forbidding everything, as RFC 9309 requires.
> Audit the homepage of example.com
https://example.com/ answered 200. Title: "Example Domain".
0 h1, 0 images without alt, 0 internal links (0 checked, 0 broken).
Sitemap: the sitemap URL answered 404.
Needs attention:
- [warning] The page has no meta description, so search engines will write their own summary of it.
- [warning] The page has no <h1>.
- [info] The page declares no canonical URL, which is how duplicate addresses for the same page get separated.
- [info] The page has no og:title or no og:image, so it will share poorly on social networks and in messaging apps.
- [info] There is no sitemap at https://example.com/sitemap.xml: the sitemap URL answered 404.
- [info] The site publishes no robots.txt. Nothing is blocked, but the sitemap cannot be declared there either.
site_crawl
Input: url — the page to start from — plus optional maxPages (25 by default,
100 at most) and maxDepth (3 by default).
Walks the site breadth-first from that page and reports what one page cannot tell you about itself: which pages share a title, and therefore compete with each other for the same search result; which share a meta description; which internal links are broken and, crucially, which page links to them; which pages still ask not to be indexed after a rebuild; and how much of the site was reachable at all.
It stays on the origin you start it on — https://example.com and
https://www.example.com are different origins, and a crawl that wandered
between them would report one site's pages as duplicates of the other's. A link
that redirects off the site is counted and left alone: robots.txt was read for
this origin and does not speak for anybody else. If the starting page itself
redirects to another origin — an apex that sends everyone to www — the report
names where it went and says to start the crawl there.
Three budgets bound it: pages, depth, and a two-minute deadline. Whichever one
ended the crawl is reported along with how many URLs were found and not visited,
because a report that does not say it saw a quarter of the site is worse than no
report. robots.txt is read first and obeyed for every URL before it is
requested; an unreadable one refuses the crawl outright, per RFC 9309.
It is deliberately not part of portfolio_report. Twenty-five pages per site
across a portfolio is a different order of cost, and
docs/adr/0021 — which
docs/adr/0010
predicted — records why it is a tool of its own rather than a depth parameter
on seo_audit.
> Crawl example.com and tell me what needs fixing
Crawled 25 pages of https://example.com, 3 levels deep.
2 broken internal links, 1 title used more than once.
Stopped at the page budget; 11 URLs were not visited.
Needs attention:
- [warning] 2 internal links are broken: https://example.com/old-pricing (it answered 404) linked from https://example.com/, https://example.com/team/ana (it answered 404) linked from https://example.com/team.
- [warning] 1 title is used by more than one page, so those pages compete with each other in search results; the widest is "Services" on 4 pages.
- [info] 6 of 25 pages have no meta description, so search engines will write their own summary of them.
- [info] The crawl stopped at the page budget with 11 URLs found and not visited, so everything here describes the part of the site that was looked at.
portfolio_report
Input: sites inline (200 at most), or file (defaults to sites.json), plus
optional checks and tags to narrow the run, client to report on one client's
sites, format (full by default; also concise, markdown, csv, ical), and
for a quarterly comparison baselineName, saveBaselineAs, includeSnapshot and
baselineSnapshot, all described below.
Runs every check across the whole portfolio with bounded concurrency and returns one report ordered by what needs action first: what is down, what expires soonest, what regressed since the last run. A site that cannot be checked becomes a finding, never a failure of the whole report.
> What needs attention across my sites this week?
3 sites checked: 0 critical, 2 warning, 0 unknown, 1 fine.
Needs action:
- [warning] Example Ltd: Plain HTTP does not redirect to HTTPS.
- [warning] Example Ltd: No Strict-Transport-Security header is sent.
- [warning] Example Net: Plain HTTP does not redirect to HTTPS.
- [warning] Example Net: No Strict-Transport-Security header is sent.
No change is reported: this server has not run a report on this portfolio before,
and the portfolio names no history file, so nothing survived the last restart.
History for this portfolio is kept in memory only. To compare across restarts,
add a "history" path to the portfolio file.
Nothing to do: Example Foundation.
Comparing across restarts. By default the previous run lives in this server process and nowhere else, so a client that restarts daily gets a comparison that never spans more than a day — and a quarter-over-quarter report cannot be produced from it at all. Add one line to the portfolio file:
{ "version": 1, "history": "upkeep-history.json", "sites": [...] }
and the run is written there instead, beside the portfolio file itself. The next
report compares against it however many restarts later, up to a hundred days. The
file names your clients and says which were broken, so it is created readable by
your account alone and replaced on every run; without that line nothing is
written at all.
docs/adr/0018 records the reasoning.
A weekly run moves that comparison forward, so a quarterly report compares against
a moment you chose instead, and the report lists what was resolved, what is new and
what is still open since then. With a history file:
portfolio_report { "saveBaselineAs": "2026-Q3" } // at the end of the quarter
portfolio_report { "baselineName": "2026-Q3" } // at the end of the next one
Sites passed inline keep no history between calls, so that no caller is ever
compared against another's run. With no file, including on the hosted instance,
which stores nothing, keep the run yourself: includeSnapshot returns it as snapshot, and baselineSnapshot passes
it back next quarter. It names your sites, so keep it as carefully as the portfolio.
Baselines last 400 days.
docs/adr/0022 has the reasoning.
Give each site a client, and a report for one of them is one call:
portfolio_report { "client": "Example Bakery Ltd", "format": "markdown" } returns a
document in plain language that says what needs a decision from them, what you are
dealing with, what changed since the baseline and how every site stands. It reads the
same way every quarter, and its dates and counts are not left to a model. It covers
what can be seen from outside, so updates, backups and traffic are yours to add.
"format": "csv" gives one row per finding for a spreadsheet, and "concise" a short
result for a large portfolio.
"format": "ical" returns a renewal calendar to import: every domain, plus certificates
within 30 days, since one renewing itself every few weeks would bury the domains. A
domain's reminder fires as far ahead as its expiryWarningDays (30 by default), a
certificate's a week ahead.
Anything already expired is an outage rather than a renewal to schedule, so it is
listed beside the calendar instead of in it, as is a domain whose registry publishes
no renewal date. Importing it again updates the events
rather than adding copies. It knows dates and
nothing else, so prices, account owners and whether auto-renew is on stay yours to note.
Each site can set maxLinks — how many internal links the seo check may
request, 0 for none. It is the setting that decides what a run costs: measured
over twenty sites, the portfolio takes about eight seconds without seo and
about a minute with it, because link checking is one request per link paced at
half a second per host.
The portfolio file format is documented in
sites.example.json. Copy it to sites.json — which is
gitignored, so a real client list never gets committed. file is resolved
against the directory the client started the server in, so give the full path
when that directory is not yours.
accessibility_audit
Input: url, plus optional standard (wcag2aa by default; also wcag2a,
wcag21aa, wcag22aa, best-practice).
Opens the page in a headless browser and runs axe-core over it. Returns the
rules that failed, how many elements failed each, CSS selectors for the first
few, and how many rules axe could not decide on its own. A page that answers
with an HTTP error is reported as unavailable — critical, audited: false —
rather than having its error page audited in its place.
This is the only tool that needs a browser, and it is optional: nothing is
downloaded when you install this project. Run npx playwright install chromium
once if you want it. Without it the tool says so and names that command, and
every other check carries on.
Automated rules find roughly a third of accessibility problems. A page with no violations passed the machine-checkable part, which is not the same as being usable — which is why the count of undecided rules is reported alongside.
Installation
Requires Node.js 22 or newer. Nothing else: installing downloads no browser,
and every check works without one except accessibility_audit.
Claude Code
claude mcp add upkeep -- npx -y upkeep-mcp
Claude Desktop
Add the server to claude_desktop_config.json:
{
"mcpServers": {
"upkeep": {
"command": "npx",
"args": ["-y", "upkeep-mcp"]
}
}
}
Restart the client and ask it to run the health tool. It answers with the
server version, the Node.js version and how long the process has been up.
A desktop client starts a server in a directory of its own choosing, usually
/. That matters for one thing only: portfolio_report and the
portfolio://sites resource look for sites.json there. Pass sites inline,
or give portfolio_report the full path — file: "/Users/you/sites.json".
From source
For development, or to run a branch:
git clone https://github.com/tiagocalado600/upkeep-mcp.git
cd upkeep-mcp
npm install
npm run build
claude mcp add upkeep -- node /absolute/path/to/upkeep-mcp/dist/index.js
The hosted instance
For anyone who cannot or would rather not run a server, there is a public one:
https://upkeep-mcp-1080119881249.europe-west1.run.app/mcp
Point any MCP client that takes a remote server URL at it — in Claude, as a custom connector. Opening the host in a browser gives a plain page saying what it is.
It is a demo. No authentication, no availability promise, no support, and it
may be switched off without notice. npx -y upkeep-mcp is the supported way to
run this, and it is what you want if these checks matter to your work.
Every tool works there, accessibility_audit included — the image ships a
browser, and every request that browser makes goes through the same rules as the
rest of the server. Someone who cannot run a server themselves should not get a
weaker tool than someone who can.
One thing genuinely differs, and it is a property of running in public rather
than a compromise: it contacts only public addresses, and only three ports —
443, 80 when checking whether plain HTTP upgrades, and 53 for the nameservers a
domain itself publishes. So it refuses to check anything on your own network,
localhost included. Use the stdio server for those.
Running your own over HTTP
The transport is the same one the hosted instance uses:
npm run build
npm run start:http -- --port 8080
The HTTP entrypoint is not the stdio one with a socket attached: a stranger is
not the person who started the process, so it applies the address and port rules
above and admits traffic through a per-caller rate limit. A caller is the
socket address unless --forwarded-for-hops says a proxy in front writes
X-Forwarded-For — 0 when it appends the caller's address last, as Cloud
Run's front end does; trusted without one, the header is the caller's to forge.
docs/deploying.md covers running it on Google Cloud Run,
and docs/adr/0012 explains the guard
and what it does not close.
Security & privacy
This server never asks for, accepts or stores credentials. It reads only information that any person with a browser or a DNS resolver could read.
- No API keys, tokens or passwords — for any service, ever. A URL with a user name or password in it is refused, and a redirect to one is not followed.
- No intrusive behaviour: no port scanning, no subdomain brute-forcing, no vulnerability probing. It inspects public configuration; it is not an offensive tool.
robots.txtis respected on any page crawl, with per-host rate limiting and an identifiableUser-Agentcarrying a contact URL.- No persistent sensitive state. Caching is in memory only, with a TTL. There is
no database. The one file this server ever writes is a portfolio run snapshot,
and only when a portfolio file names a
historypath for it. - The only third parties this server contacts are the ones that hold the
answer: the registry's own RDAP server, IANA's RDAP bootstrap file,
cloudflare-dns.comfor whatnode:dnscannot ask (whether a DNSSEC delegation is signed and validates, and CAA when the system resolver returns none), a certificate's OCSP responder, and Let's Encrypt for the renewal window of a certificate it issued. The exception isaccessibility_audit: its browser loads the page as a visitor would, so whatever the page embeds — scripts, fonts, analytics — is fetched from wherever the page names.SECURITY.mdlists them and what each one learns.
Limitations
Stated plainly, because a tool that hides what it cannot do is worse than one that does less.
- Certificate revocation lists are not downloaded. Revocation is checked over
OCSP, and only over OCSP. A CRL is a file of every serial an authority has ever
revoked — megabytes, fetched to answer one question about one certificate — so
a certificate whose issuer publishes no responder is reported as unchecked,
with the reason, rather than being judged from a file this tool declined to
read. Since 2025 that covers Let's Encrypt and Google Trust Services, which is
a large share of the web. See
docs/adr/0017. - An OCSP answer is a point in time, not a subscription. Responders pre-sign
about a week ahead, so
goodmeans "not revoked as ofthisUpdate", and a certificate revoked an hour ago may still read as good until the authority publishes its next answer.producedAt,thisUpdateandnextUpdateare all reported so the age of the answer is visible rather than implied. - Some registries publish no expiry date.
.de,.nl,.no,.auand.fido not publish one over any protocol. The result names the registry and says so, instead of showing an indefinite "unknown". Registration data comes from RDAP only; there is no WHOIS fallback, anddocs/adr/0004explains why. - DNSSEC is checked through a resolver, not verified here. The tool reports whether a delegation is signed and what Cloudflare's validating resolver makes of it, which is enough to catch a broken chain. It verifies no signature itself.
- The parent's delegation is not compared with the zone's own. "The nameservers at the registrar are not the nameservers in the zone" is the other classic delegation fault, and answering it means querying the parent zone's servers for a referral — a second hop and a different feature. What is checked is whether the zone's own nameservers agree with each other.
- Response time includes connection setup. It is wall clock to the first response headers, covering DNS, TCP and TLS, so it is not a measure of server processing time.
seo_auditaudits one page, not a site. It requests the page's internal links to find broken ones, but it does not crawl: there is no second level. Auditing a site means calling it for the pages that matter.- The sitemap is checked against the protocol's rules, not against its XSD.
It establishes that the document exists, declares
<urlset>or<sitemapindex>, how many<loc>entries it holds, and whether it arrived gzipped — asitemap.xml.gzis unpacked before it is read, capped at 16 MiB so that a decompression bomb is refused rather than unpacked. It then checks the rules that decide whether a consumer keeps an entry: the namespace,<loc>present, absolute, escaped, within 2048 characters and on the sitemap's own host,<lastmod>a W3C Datetime,<changefreq>and<priority>within their ranges, and the 50,000-entry limit. There is no XML parser and no schema validator here, so four rules are deliberately left unchecked — the 50 MB size limit, an index listing another index, duplicate entries, and whether a<lastmod>is true.docs/adr/0019lists them and says why. The rules are checked against what was read, so a sitemap past the read limit is judged on the entries before the cut and the report says it was truncated — nodejs.org's stray entries on another host sit past it, and are not reported. - A page nested hundreds of levels deep is refused, not audited. HTML
parsing costs roughly the square of the nesting depth, misnested tags can
make it walk the same long lists again and again, and so can one tag with tens
of thousands of attributes, so a document built to be absurd would block the
server for minutes. The parser stops past 512 levels or past a fixed budget of
work — far beyond any real page — and
seo_auditreports the refusal. - Accessibility is only checked as far as a machine can. Automated rules find roughly a third of real problems. The tool reports what axe could not decide rather than counting it as a pass, but no green result here is an accessibility statement.
- Nothing here judges how a page ranks.
seo_auditreports what is in the HTML. Rankings depend on things no public endpoint exposes. - "What changed since last time" lasts as long as the server process, unless
you ask for otherwise. For a portfolio file the previous run is held in memory and
never written anywhere, so a restarted server — which for a desktop MCP client
is a daily event — has nothing to compare against, and says so rather than
implying nothing changed. Adding a
historypath to the portfolio file makes the comparison survive a restart, for up to a hundred days; nothing is written without it. Sites are compared only on the checks both runs measured, so a quick uptime-only pass never invents regressions in the run after it.docs/adr/0011anddocs/adr/0018explain the trade. - Only the previous run and the baselines you name are kept, never a series.
Each run replaces the last, except that a narrower run does not erase a wider one:
an uptime-only pass or a
tagsrun leaves the fuller outcomes in place. - Certificates and domains are judged on different clocks. A registration is
a warning inside 30 days, or the site's
expiryWarningDays, which applies to the registration only. A certificate is judged on its own lifetime: a 90-day one inside 14 days, a 45-day one inside 7, a 6-day one inside 1, and a long one at most 30 ahead. ACME clients renew at about two thirds of a certificate's life, so warning earlier would fire on nearly every healthy site.
Documentation
CONTRIBUTING.md— development setup, and how to add a tooldocs/architecture.md— how a request flows, and how it is tested without a networkdocs/prior-art.md— the MCP servers that already do parts of this, what they do better, and the gap this one fillsdocs/deploying.md— running the HTTP instance, and what is different about itdocs/adr/— one short record per structural decision, including why the browser is optionalSECURITY.md— threat model and reporting policyCHANGELOG.md
License
MIT — see LICENSE.
upkeep-mcpnpm install upkeep-mcp