ProductIntegrations
Your AI SRE finds the cause. Observer proves the impact.
Bring the alerts and AI tools you already run. Observer turns them into measured status, an evidence trail and customer updates a person publishes.
In
What you already run
- Prometheus Alertmanager
- Grafana Alerting
- Datadog
- AWS CloudWatch
- Any JSON sender
- Your AI SRE agent
Observer
One measured record
- A status per metric
- The incident and its impact window
- Customers and SLOs affected
- Notes from other tools, kept internal
Out
What customers and auditors see
- Status pages, per customer
- Updates drafted from the evidence
- Signed SLA statements
- Alerts to 10 channels
Where it fits
Two jobs, kept apart.
Your AI SRE
Finds the cause.
Reads logs and traces, forms a hypothesis, suggests the fix. Its findings land on the incident in Observer as internal notes, next to the measurements.
Observer
Proves the impact, and tells customers.
Measures who was affected and for how long, keeps the timeline, drafts the customer update from that evidence, and signs the SLA statement at the end of the period. A person publishes.
Alerts in
Your alerts drive status, with rules you can read.
Senders
Alertmanager, Grafana, Datadog, CloudWatch, or any JSON.
Point an Alertmanager webhook, a Grafana contact point, a Datadog webhook or a CloudWatch alarm's SNS topic at your integration's URL. Anything else can send Observer's documented JSON.
Routing
Prometheus-style matchers, first rule wins.
Each rule matches on labels and sets the status it drives: unhealthy, degraded, or by severity. An alert with a severity Observer cannot classify reads unhealthy, never healthier than it is.
Security
A secret per integration, never in the URL.
Each integration has its own secret, sent as a header and stored only as a hash. Grafana and JSON senders can also sign each request. Each integration takes up to 120 requests a minute.
Plans
From Starter.
3 inbound integrations on Starter, 10 on Pro, 25 on Business and unlimited on Enterprise.
AI tools
Findings on the record, never on the page.
Notes
Other tools write to the incident's internal timeline.
An AI SRE agent or a runbook bot posts its findings to the incident with an API key. Notes stay internal: no status page, email, webhook or subscriber ever sees one.
Evidence
And read the measured impact back.
The same API, and the MCP server for AI assistants, return the incident's evidence: the impact window, the checks and customers affected, SLO burn and data coverage.
Updates
Customer updates come from the evidence.
Observer drafts the customer update from what it measured, never from a guessed cause, and a person on your team publishes it.
Alerts out
And out to the channels your team works in.
Slack, Microsoft Teams, PagerDuty, Jira, email, signed webhooks and four more, each with its own routing. Recoveries always get through.