--:--:-- utc·replaying incident #4127 — recorded last night
Beacontrains a support chatbot on your website, docs and FAQ so it answers customers all day with citations — then wires it to your uptime monitors so it never lies when you’re down. This page is not a pitch. It is the tape of the 37 minutes that make the pitch unnecessary.
live demo feed — these numbers are moving because monitors never sleep
t+0:00 · 23:04:31
A 10-second timeout where 120 milliseconds should be. Beacon does not page you — one blip is just the internet being the internet. It schedules an immediate re-check and holds its breath.
avg 118ms · api /v1/health
TIMEOUT · 10,000ms
t+0:31 · 23:05:02
The second failed check confirms it. In the same second, Beacondoes four things: pages you, opens a public incident, re-briefs the chatbot, and flips the status page. You haven’t unlocked your phone yet.
every incident leaves a paper trail ↗
beacon incident record
#4127 — api /v1/health
23:04:31check missed — api /v1/health (timeout 10s)
23:05:02second miss. threshold met → incident #4127 opened
23:05:02email dispatched → ops@acme.dev
23:05:04chatbot re-briefed: outage-aware answers ON
23:05:06status page flipped to “degraded”
✉ dispatch · 23:05:02 → ops@acme.dev
[beacon] DOWN: api /v1/health — incident #4127 opened. 2 failed checks. Status page updated automatically.
t+1:09 · 23:05:40
Beacon trains on your website, docs and FAQ, so it answers everyday questions with citations — try “how do I export my data?” But it also checks live monitor state before every reply. Ask an outage question, then flip the switch to status-blind to hear what every other chatbot would have said at 23:05 last night.
← the blind answers are how 214 tickets happen
acme.dev · support
Hi, I'm the acme.dev support bot — trained on acme.dev's docs & FAQ, and wired to the live outage above. Ask me an everyday question or an outage one.
t+36:46 · 23:41:17
Two clean checks close the incident, the status page goes green, subscribers get the all-clear — and the wall remembers everything for 90 days, in public.
0 min
detected → resolved
0
visitors told the truth by the bot
0
support tickets created
marketing-site
app.acme.dev
api /v1/health
← 60 days ago · the one red cell is tonight · today →
the system
01 · sensors
HTTP checks every 30s–5min from the moment you paste a URL. Latency, status code, keyword match.
02 · reflex
Two misses opens an incident, pages you, and starts a public timeline. Two clean checks closes it.
03 · voice
The bot is trained on your site, docs & FAQ to answer everyday questions with citations — and both read live monitor state, so mid-outage customers get the truth with a link, not a shrug.
the connection nobody else has: the mouth is wired to the pulse
pricing
# beacon.config — pro
plan = "pro"
price = $49/mo
ai_chatbots = 3 × 500 pages
messages_per_mo = 3,000
answer_model = claude-haiku-4-5
monitors = 20 @ 60s checks
status_pages = 3
branding = off
$49/mo
For a live SaaS. Sixty-second checks, three thousand answers a month, your branding only.
Deploy proupgrade or bail anytime · billing by stripe