Error monitoring · APM · uptime · on-call
The whole incident,
not a slice of it.
One product that catches the exception, holds the trace, runs the probe that agrees with it, and pages the person on call — on one meter, at the edge, and reachable by your agents as tools rather than as a screen.
100k events/mo free · no card · unlimited viewers on every planincident INC-22911m 51s to ack
- 14:02:11.204error TypeError: cannot read properties of undefined (reading 'total')
- 14:02:11.204transaction POST /checkout/confirm — p99 2.41s, db.query 1.98s
- 14:02:11.331issue grouped into CHK-418 · 1 → 340 events in 8s · regression
- 14:02:19.000monitor checkout-api https · 3 of 3 regions agree DOWN
- 14:02:19.412incident INC-2291 opened sev3 · policy payments-primary
- 14:02:19.480notify push + slack + pagerduty → rae (on call, schedule eu-week)
- 14:04:02.117incident acknowledged by rae · escalation to tier 2 cancelled
Illustrative incident. Field names, id formats and statuses are the ones the product emits.
Migrating
Change one line
We speak the Sentry envelope protocol on ingest. Your existing SDKs, releases, source maps and breadcrumbs keep working — you are changing a hostname, not an instrumentation strategy.
Sentry.init({- dsn: "https://<key>@o0.ingest.sentry.io/<project>",+ dsn: "https://<key>@ingest.bugwatch.io/<project>",tracesSampleRate: 0.2,});
no rewrite
No SDK swap and no vendor-specific API calls in your code.
dual-send
Send to both while you evaluate, then cut over when you are ready.
backlog
Import triaged issues so a year of decisions is not reset in week one.
Agent parity
If your agent cannot do it, it is not done
Every dashboard page has a matching MCP tool, generated from one registry — so a tool can never carry weaker auth or looser validation than the route it wraps. Same issue, three ways in:
dashboard
/o/acme/p/web/issues/CHK-418rest
GET /v1/orgs/acme/issues?status=unresolved&project=webmcp
search_issues(org="acme", project="web", status="unresolved")one registry
The API executes a tool by dispatching its REST mapping through its own handlers. A parity test fails the build when a route ships without a tool.
docs as data
Every page is served as HTML and as raw Markdown, indexed in
/llms.txt, and readable over MCP.audited
Every mutation an agent makes is written to an audit feed you can read back.
What you get
Five signals, one meter, one timeline
errors
Error monitoring — Grouped issues with the stack frames that matter, not a wall of duplicates.
performance
Performance — Transaction traces and web vitals, sampled honestly.
uptime
Uptime monitoring — HTTP, TCP, TLS, DNS, and heartbeat checks from multiple regions.
oncall
On-call and escalation — Schedules, rotations, and escalation chains that reach a human.
Pricing
One meter. No seat tax.
Free
Free
100k events/moStarter
$29
1M events/moTeam
$79
5M events/moBusiness
$249
20M events/moScale
$499
50M events/moCreate an accountHard cap by default, so nothing bills you by surprise.