Beamfar
Sign inStart the 14-day trial

Last updated

What Beamfar watches, and how it proves it.

Beamfar checks each client site from two regions and writes down what it saw. Every part on this page exists today, with the facts behind it. What is not built is listed at the bottom.

Start the 14-day trialCheck a site now
  • Nothing to install on client sites
  • Any stack
  • No card for the trial

One site, read from two regions

www.birch.example

  • Page200 in 240 msSteady
  • Certificate62 days leftValid
  • DNSA 198.51.100.4Changed 11:52
  • Redirect301 → /en/Same
  • Searchindex, followIndexable
  • Domainuntil 14 Jan 2027Registered
Demo data, made up: what the checks of one site read.

Know it is down, with the proof

A site is down only when two regions agree. Every alert carries what each region saw and what changed just before it.

Down only when two regions agree

A site is called down when two regions both fail to get a valid answer. If only one region fails it is shown as a regional issue and sends no alert.

Downboth regions agree: one alert

  • A check with no valid result is shown as data unavailable, never as up and never as zero.
  • Every alert lists what each region saw, so you can forward the evidence to the client or the host.
How the two-region rule works

A timeline of what changed before the incident

Between checks, Beamfar records changes to DNS records, the certificate, the final redirect address and the SEO signals of the page.

  1. DNS: an A record changed
  2. Region A: no answer
  3. Both regions fail: down
  • When an incident follows within the next hour, the change is marked next to it.
  • The question after an outage is usually “what changed?”. The answer is already on the page.

Why it went down, and what to tell the client

When two regions confirm an outage, the incident page shows the likely cause with its confidence, the evidence behind it, who has to act and the fix steps.

Likely cause
The hosting provider Medium
Ruled out
Certificate: valid
Not checked
Provider status page: no feed

Demo data, made up.

  • It lists what was ruled out and what could not be checked. When the evidence points nowhere, it says unknown.
  • A likely cause is a ranked reading of stored evidence, not a diagnosis, and no AI model decides it.
  • Message your client writes the note for you in three lengths and as a status page update, in plain words and without blame. You edit it before it goes anywhere.

Show the client the work

A monthly report, a public status page and a read-only portal for each client, every number from a stored check. These are real screens of the app, on made-up demo data.

Demo agency / Reports / October 2026Demo data, made up
The monthly report of a demo client: a summary of checks, confirmed incidents and changes, then recommended actions.

A monthly report per client

One report per client, every number taken from stored checks, with plain-language explanations of incidents. It prints to PDF.

  • On the Agency and Agency Plus plans your own name, logo and accent colour are on it.
  • It carries the real-user speed of each site too.
  • See a sample monthly report (demo data, made-up sites).

A public status page per client

A page your client can open. Only confirmed incidents appear on it: two regions for page, API, port and blocklist checks, the ping receiver for scheduled tasks. A day without data is never shown as green.

  • Plans include 1, 10, 30 or 60 pages (Solo, Team, Agency, Agency Plus; 1 on the Free plan). With the Agency and Agency Plus plans the page carries your brand.
  • Visitors can subscribe by e-mail, with a double opt-in. They get each confirmed incident, its resolution and the updates you post under it (Investigating, Identified, Monitoring, Resolved), and leave with one click.
  • See a sample status page (demo data, made-up sites).

A read-only portal for each client

Invite a contact of your client to see that client’s sites, incidents, status pages and monthly reports, and nothing of your other clients.

  • They change nothing and take no seat of your team. The limit is enforced by the server and by the database, not only by the screen.
  • On the Agency and Agency Plus plans the portal carries your name, logo and colour. The client portal is part of the Team, Agency and Agency Plus plans.
How to invite a client

Catch what “up” hides

An uptime check sees that the page answered. Beamfar also reads the certificate, the DNS records, the search settings, the domain and what a visitor would see.

  • SEO-risk alerts

    An uptime check says up when a page answers. It cannot see that a deploy added noindex.

    1. A deploy changes the robots tag. The site is up.
    2. First sighting: on the timeline, no alert yet
    3. A second check confirms it: SEO-risk alert, with the tag that changed
    • Beamfar reads the robots meta tag and header, robots.txt and the canonical address, and alerts when a page becomes noindex, robots.txt starts blocking it, or the canonical moves to another host.
    • Alerts are sent after a second sighting. It reads the pages you add as monitored URLs; a page that is already noindex when you add it is listed in the first-day findings, and alerts start from the first change.
    • Tested on test pages; not yet run on customers’ sites.
    Why this matters after a deploy (draft article)

  • Certificate watch

    The TLS certificate of each site is read on every check: who issued it, when it expires, and when it is replaced.

    Certificate expires in 9 days

    A certificate that is about to expire or fails verification shows up before visitors see a browser warning.

  • Domain registration expiry

    The registration date of each domain is looked up through the public RDAP service, at most once a day per domain, and an alert is sent as the date gets close.

    Domain expires 14 Jan 2027

    Where a domain ending does not publish the date, the page says data unavailable.

  • DNS records and record changes

    Address records (A, and AAAA when the name has them) are read from two public resolvers and compared with the last answer; a change is recorded with its time.

    Demo data, made up. AAAA records: changed, yesterday.

    • The free site check also shows alias, name server and mail records once.
    • A DNS check also watches the name servers, mail servers (MX), SPF, DMARC and CAA records of the name, about every 15 minutes. A change counts only when two regions see it, or one region sees it twice, and the alert says which records were added and which were removed. The records are only read: mail delivery is not tested.

  • Site security and SEO integrity

    Each site is read from outside, the way a visitor sees it: scripts and frames from a new host, links and text a visitor cannot see, spam vocabulary, redirects to another site, software versions on a list of exploited flaws and missing security headers.

    • Every finding is a possible problem with its evidence, not a verdict that the site was hacked.
    • Two kinds of check run only after you prove with a DNS TXT record that you control the domain: the crawler views (the first pages as Googlebot, as a visitor from a search result and as four AI crawlers) and the active checks for files that should never be public, such as a Git folder or an .env file.
    • Beamfar is not a firewall: it cannot block an attack or clean a site. Tested on test pages; not yet run on customers’ sites.
    What the security checks read

  • Real-user speed (Core Web Vitals)

    What real Chrome visitors measured on each site over the last 28 days, from Google’s Chrome UX Report: loading (LCP), responsiveness (INP) and visual stability (CLS), each rated by Google’s own thresholds.

    Loading (LCP)
    2.1 s
    Good
    Responsiveness (INP)
    260 ms
    Needs improvement
    Stability (CLS)
    0.04
    Good
    • It is looked up once a day for the whole site and for its home page, and it goes into the monthly report.
    • A site with too few Chrome visits says “not enough real-user data” instead of a figure, and the Client Risk Radar lists a site whose loading got slower for real visitors. This is field data from visitors, not a lab test.
    • Tested against a stand-in of Google’s API; the live lookups start once the service is connected to it.
    Where the figures come from

  • Client Risk Radar

    One list across all your clients of what could hurt a site next: certificates and domains close to expiry, registry flags and registrar transfers, DNS changes, search blocks, software on the list of exploited flaws, and a host that carries too many of your sites.

    • CriticalCertificate of client-one.example: 5 days left
    • WarningDomain of client-two.example: 21 days left
    • NoticeOne host carries 7 of your sites
    • Each risk carries the evidence it comes from.
    • An opt-in e-mail brings the list on Mondays, only in a week with something to look at. A check that could not be made is listed as such, never as an all-clear.

Alerts, and the work around them

Alerts go where your team already works, someone takes each one, and planned work never wakes anybody.

Alerts where you already work

Alerts go by email, to Slack, Microsoft Teams, Discord, Telegram, Google Chat, PagerDuty, Opsgenie or ntfy, or to a signed webhook.

  • Email
  • Slack
  • Microsoft Teams
  • Discord
  • Telegram
  • Google Chat
  • PagerDuty
  • Opsgenie
  • ntfy
  • Signed webhook: Team and up
  • The signed webhook is on Team and up; every other channel is on every paid plan, and the Free plan alerts by e-mail and push.
  • Destinations, tokens and signing secrets are stored sealed, and a channel gets alerts only after its test message was accepted.
  • A channel can belong to one client, so the client’s own Slack, Teams or address hears about that client’s sites only.
  • The seven newer channels need no paid add-on; they were tested against each service’s documented formats, not yet against the services themselves.

Acknowledge, then escalate

Acknowledge an incident from the app, the API or the link in the alert, so the team sees that someone is on it; Slack, Teams, Telegram, Google Chat and ntfy show it as a button.

  1. Outage confirmed by two regions: alert sent
  2. Not acknowledged after 15 minutes: escalated once
  3. Acknowledged with the Slack button
  4. The site answers again: resolved
  • The link works for that incident only and asks for one click, so a mail scanner acknowledges nothing.
  • If nobody acknowledges an incident within the minutes you set, from 5 to 120, a second channel is told once: never during a maintenance window, never after someone acknowledged it and never for a resolved incident.
  • Acknowledging is on every plan; escalation is on the Team, Agency and Agency Plus plans.
How acknowledging and escalation work

Maintenance windows

Schedule planned work for the whole organisation, one client or one site, once or every week in your own time zone.

Planned Every Sunday, 02:00, Europe/Bucharest (demo)

  • Checks keep running, but during the window alerts are held back, no incident opens, and the time is left out of uptime and the monthly report.
  • A window marked public shows on the client’s status page as planned maintenance.
  • A window cannot start more than five minutes in the past, and one that has ended cannot be changed, so nobody can draw a window around an outage afterwards.

Built for a portfolio

Paste a list of sites: both regions start checking as soon as you add them. Group sites by client and invite people with four different roles.

Demo data, made up: four sites pasted at once, then grouped by client.

  • Owner, admin, member and viewer roles.
  • Export your data as JSON and CSV at any time.
  • Plan limits: 10, 40, 100 or 200 client sites.

Also built

The rest of what it can do, one sentence or two each.

Beam, the assistant
On this site it answers from our own pages and links the page each answer comes from, or says it does not know and offers a person.
  • In the app it reads your own sites, incidents and reports and changes nothing.
  • Each answer says whether it was matched from the documentation or written by an AI model and checked against it; in the app the AI model is used only on paid plans.
Sign-in
With a password, a link sent by e-mail, or a Google or GitHub account; owners and admins also use a second factor.
  • Google and GitHub sign-in were tested against stand-ins, not yet against Google and GitHub themselves.
An API and tools for an AI assistant
From the Team plan: a REST API and twelve read-only tools over MCP, with keys that act as the member who made them; keys read on the Team plan and can also write on the Agency and Agency Plus plans.
Scheduled-task, API, port and blocklist checks
The plan sets how many.
  • Scheduled-task (cron) monitors with a secret ping address
  • API checks with assertions on the status, headers, JSON values and response time
  • TCP port and reachability checks, and DNS blocklist checks
  • Response times per region with a slowdown notice
Site crawl and sitemap
A plan-limited crawl, daily or weekly, reads a site’s own pages and reports broken pages and links, redirect loops and mixed content; the sitemap is found, checked, counted and sampled.
Status pages on your own domain
On the Team, Agency and Agency Plus plans a status page can be shown on a host name you verify with a DNS record, served with a certificate. Built and tested offline; not yet tested with real DNS and TLS.
Status badges and a widget
An opt-in badge for a site or a status page, with the status and the measured uptime: an image, a text widget without a frame, or JSON, at an address nobody can guess. Remove it and the address stops working. On every plan.
An installable app with push notifications
Install the app from the browser and turn on incident notifications for your own devices.
  • They travel through the browsers’ own push services, encrypted for your browser, with no app store and no paid service.
  • Tested with a stand-in push service, not yet on a real phone.
A free site check and free tools
Without an account, anyone can run the free site check or one of the five free tools: certificate, domain expiry, indexability, DNS lookup and security headers. Nothing about the address is stored.

Planned, not built

These are not available, and the product does not pretend otherwise.

Not built yet
Payments: none are taken yet; paid plans open soon. SMS or phone alerts.
Not planned as checks
Lighthouse or browser performance audits (the real-user speed figures come from Google’s Chrome UX Report), and server, network or application monitoring. For those, a broader tool is the better choice. The comparison pages say where.

Questions

What does it check on a client site?

Beamfar checks the page, the TLS certificate, DNS records and their changes, redirects, SEO-risk signals (noindex, robots.txt blocking, canonical moved to another host) and domain registration expiry, and it reads the site from outside for signs of a security problem.

How is a site called down?

Only when two regions agree. One region failing is a regional issue and sends no alert.

Do I need to install anything on client sites?

No. Checks run from outside, so any stack works.

What is not built yet?

Real payments and SMS or phone alerts. Status pages on your own domain are built (Team, Agency and Agency Plus plans) but not yet tested with real DNS and TLS.

Guides that go with this page

Longer reads on the same ground, in plain words, each with its sources.

Know before your clients do.

Add your client sites and both regions start checking. Judge the evidence yourself.

Start the 14-day trialCheck a site now

No card for the trial.