Barks only when it’s real.

Keenbark watches your clients’ websites from the outside. It is built to raise the alarm only when two independent observers agree — and when it couldn’t look, to say so instead of showing green.

Ask for early accessFor web agencies. Not open yet.

One row per client site. Every column judged on its own.

Each row’s state is the worst of its columns: no blended score, no green by default. Each reading says how old it is, and a missing reading says why.

An example of the Dashboard’s Websites list, with four of its six columns, not a live reading.
WebsiteReachabilityCertificateDomain registrationCritical page
bakery-client.com

Verified

Not covered: the flow behind “Book a table”, end to end.

Reachable as of 38 seconds ago
Valid until 18 Nov 2026
Registered until 31 Aug 2027
Renders as declared “Book a table” is visible
dental-studio.fr

A possible problem, being verified

A possible problem the server answered HTTP 503, a server-side failure (read 02:14 UTC) — being verified, as of 40 seconds ago
Valid lapses in 9 days — nothing has failed yet
Registered until 3 Feb 2027
Not covered none is declared
hotel-riviera.md

Not verified

Not covered: the flow behind “Check availability”, end to end.

Not known connected, and no response within the 30-second observation budget; slow and alive is not unreachable, as of 2 minutes ago
Valid until 2 Jan 2027
Not covered the registry for .md publishes no RDAP service, so this release cannot read its date
Renders as declared “Check availability” is visible
  1. One observer saw dental-studio.fr fail. The Dashboard shows it as a possible problem; your inbox waits until an observer in an independent place sees it fail too.
  2. hotel-riviera.md took the connection and sent nothing back within 30 seconds. Slow is not down, so the cell says Not known, what was read, and how long ago: never a green it did not check.

Live reading keenbark.com, as Keenbark read it at 11:36 UTC: Reachable as of 54 seconds ago — the homepage only. No other page or capability is covered.

For now its two observers run on one machine, so they count as one witness — the reading says so itself, under “How this was judged”.

How this was judged

What it asked, how sure it had to be, and how often and from where it looked, in the product’s own words:

Scope:
the homepage only. No other page or capability is covered
Questions asked:
1 expectation
Importance:
inferred by the platform, not declared by you; its weight is not yet calibrated
Evidentiary threshold:
in force, provisionally: consecutive readings across independent domains, not yet measured on the Websites watched here
Observation intensity:
2 vantage points, realised at least every 5 minutes, they share a failure domain, so for a failure in either they are one observation
How answers are evaluated:
judged against a provisional budget the platform set (5 minutes), not one declared for this dimension; which observers count as independent, and whether each was itself working
Recorded as:
exp-c3-reachability.v1 · weights-uncalibrated · threshold.C3@2 · independence.C3.v2 · policy-c3-reachability.v1 · freshness.provisional.C3@1 · observer-health.control-fetch.v1
  • Freshness is judged against a provisional platform-declared budget for this dimension.
  • Whether an intermediary is serving this from cache over a failed origin is not determined. We see the response, not the origin.
  • A homepage that answers slowly, within the time allowed, still reads as reachable. Response times are recorded and compared with nothing, so gradual slowing is not detected.

Not covered: Certificate revocation (no outside observer can see it); DNS (this release does not observe it); Gradual slowing — no normal speed is learned, so slow but usable is not detected; Content: what the homepage should say — nothing is declared or learned; The flow behind it, end to end (booking, checkout) (observing it would affect your system, and nobody has consented). Cannot observe: Anything behind a login — the platform holds no credentials.

What it watches, from outside

Does the site answer?
Reachability. The homepage, asked from every vantage point.
Is it the real site?
Response shape. A success, HTML, on the site’s own domain — not a redirect to somewhere else.
Is it a broken page that still answers?
Error signatures. A WordPress database error, a crashed PHP or framework page, a server’s default page, a parked domain.
Will the certificate lapse?
Certificate. Chain, name and expiry: warned about before it lapses, never called an outage.
Will the domain lapse?
Domain registration. When the registration runs out, read from the registry itself where it publishes one, as .com, .org, .uk, .fr and .nl do. Many country registries publish none, among them .eu, .de, .it, .ro, .md and .io; for those the list says Not covered, and why.
Does the page that earns money work?
The page that matters. One page you name, such as a checkout or a booking form, loaded in a real browser, shut in a container of its own.
  • Does the site answer?

    Reachability. The homepage, asked from every vantage point.

  • Is it the real site?

    Response shape. A success, HTML, on the site’s own domain — not a redirect to somewhere else.

  • Is it a broken page that still answers?

    Error signatures. A WordPress database error, a crashed PHP or framework page, a server’s default page, a parked domain.

  • Will the certificate lapse?

    Certificate. Chain, name and expiry: warned about before it lapses, never called an outage.

  • Will the domain lapse?

    Domain registration. When the registration runs out, read from the registry itself where it publishes one, as .com, .org, .uk, .fr and .nl do. Many country registries publish none, among them .eu, .de, .it, .ro, .md and .io; for those the list says Not covered, and why.

  • Does the page that earns money work?

    The page that matters. One page you name, such as a checkout or a booking form, loaded in a real browser, shut in a container of its own.

And it says what it does not watch: DNS, revocation, gradual slowing, what the homepage should say, anything behind a login, the flow behind a form.

A night with Keen

An example of how Keenbark is built to behave, not a live reading. Two observers, in two independent places, look at dental-studio.fr once a minute, at the same moment. Today Keenbark’s own two observers share one machine, so it confirms nothing yet: a night like this needs the second place.

  1. From the outside, everything Keenbark observes is normal.

    Not emailed

  2. One observer gets HTTP 503 from the site; the other gets the page.

    The Dashboard shows “a possible problem”. Two minutes later both observers reach it, and it is recorded as not confirmed: one witness is no witness.

    Not emailed

  3. One observer fails its own check of a neutral page.

    Its readings are set aside and marked “on our side”: that is our network, not your client’s site.

    Not emailed

  4. Neither observer can reach the site, two readings in a row.

    One email to your agency

    dental-studio.fr: Confirmed — Reachability, the homepage only. First failing reading 03:40 UTC; confirmed 03:41 UTC.

    The email, word for word

    That email in full, as Keenbark writes it. It says no more than the Dashboard would, and when the site answers again one more email says so in the same form.

    From Keenbark To your agency Subject dental-studio.fr: Confirmed — Reachability, the homepage only. First failing reading 03:40 UTC; confirmed 03:41 UTC.

    dental-studio.fr: Confirmed — Reachability, the homepage only. First failing reading 03:40 UTC; confirmed 03:41 UTC.

    Broken as of 03:41 UTC — the homepage only. No other page or capability is covered.

    How it was judged
    Scope:
    the homepage only. No other page or capability is covered
    Questions asked:
    1 expectation
    Importance:
    inferred by the platform, not declared by you; its weight is not yet calibrated
    Evidentiary threshold:
    in force: consecutive readings across independent domains
    Observation intensity:
    2 vantage points, realised at least every 60 seconds, 2 independent failure domains
    How answers are evaluated:
    judged against a provisional budget the platform set (5 minutes), not one declared for this dimension; which observers count as independent, and whether each was itself working
    Recorded as:
    exp-c3-reachability.v1 · weights-uncalibrated · threshold.C3@2 · independence.C3.v2 · policy-c3-reachability.v1 · freshness.provisional.C3@1 · observer-health.control-fetch.v1
    • Freshness is judged against a provisional platform-declared budget for this dimension.
    • Part of the problem confirmed at 03:41 UTC, whose first failing reading was at 03:40 UTC.

    Not covered: Certificate revocation (no outside observer can see it); DNS (this release does not observe it); Gradual slowing — no normal speed is learned, so slow but usable is not detected; Content: what the homepage should say — nothing is declared or learned; A critical page in a real browser — none is declared (waiting on a person's decision); The flow behind it, end to end (booking, checkout) (observing it would affect your system, and nobody has consented). Cannot observe: Anything behind a login — the platform holds no credentials.

    1. The subject line names the Website first, then what failed, its first failing reading, and when it was confirmed.
    2. Then the state, as of a clock time that stays true however late it is read, and exactly what it covers: the homepage, nothing more.
    3. How it was judged, in the email itself: what was asked, how sure it had to be, how often and from where it looked, the names it was recorded under, and what it could not tell.

    If an email ever proves wrong, a correction follows the same way, quoting what it withdraws.

  5. Both observers reach it again, and it holds.

    One email to your agency

    dental-studio.fr: Recovered — Reachability, the homepage only: the problem confirmed at 03:41 UTC, answering again since 04:05 UTC.

Two emails all night. Both were real.

Ask for early access

Before you ask

How am I told?
By email, once two observers agree, and once more when the site answers again. A possible problem stays on the Dashboard, not in your inbox.
Why wait for a second observer?
One witness is no witness: a failing network on our side is not your client’s site being down. The wait is a minute or two, and it is what spares you the false alarm at three in the morning.
Can my clients see it?
If you let them. Each client can have a read-only view of their own sites: the same states, explained in words written for them, every confirmed problem of the last 25 months, a copy of it they can take themselves, and nothing of your other clients.
And when Keenbark is the one that fails?
It says so, on the Dashboard. A reading it could not take is marked “on our side”, never shown as green, and never emailed as your client’s problem. If it stops looking altogether (a full disk, a clock gone wrong, a record it cannot read), every page says so, and the time it missed stays on record as a gap on our side, not as a quiet night. It does not email you about its own trouble.
What does it cost?
Not set yet. Early access is how we learn what it is worth to an agency.
When does it open?
There is no date yet. Agencies on the early-access list hear first.
  • How am I told?

    By email, once two observers agree, and once more when the site answers again. A possible problem stays on the Dashboard, not in your inbox.

  • Why wait for a second observer?

    One witness is no witness: a failing network on our side is not your client’s site being down. The wait is a minute or two, and it is what spares you the false alarm at three in the morning.

  • Can my clients see it?

    If you let them. Each client can have a read-only view of their own sites: the same states, explained in words written for them, every confirmed problem of the last 25 months, a copy of it they can take themselves, and nothing of your other clients.

  • And when Keenbark is the one that fails?

    It says so, on the Dashboard. A reading it could not take is marked “on our side”, never shown as green, and never emailed as your client’s problem. If it stops looking altogether (a full disk, a clock gone wrong, a record it cannot read), every page says so, and the time it missed stays on record as a gap on our side, not as a quiet night. It does not email you about its own trouble.

  • What does it cost?

    Not set yet. Early access is how we learn what it is worth to an agency.

  • When does it open?

    There is no date yet. Agencies on the early-access list hear first.

Early access, for web agencies

Keenbark is not open yet. If you look after many client sites, leave your address: we will write when there is a place, and ask what you would want from it.

Enter your work email. Check the address. A full address, with a dot in the part after the @, like you@agency.com.

Tell us more optional

We use this only to write to you about Keenbark. What we keep, and for how long.