Incidents integration
Let an agent raise the alarm without letting it raise forty. It opens incidents under a daily ceiling, and resolving one can wait for a person, so an incident does not go quiet on an agent's say-so.
Connect
bound to api.pagerduty.com
What you create
Connecting allows nothing by itself. No API returns the credential afterwards, to you, to the console or to an agent, and Anlyon refuses to send it anywhere but api.pagerduty.com.
Operations
Each one is a named tool, off until you switch it on, and Anlyon makes every call with the credential your agent never sees. Reads return what PagerDuty answers and change nothing. Writes count against a limit you set, can wait for a person, and are read back from PagerDuty wherever it allows.
Reads
pagerduty-list-incidents
List incidents, newest first, 25 per page: title, status, urgency, service and assignees. Without since/until PagerDuty searches the last month, and date_range all searches every date; filter by up to two statuses (pass triggered and acknowledged for every open incident), a service, an assigned user or an urgency. When the response says more is true, pass offset 25, 50 and so on for the next page.
pagerduty-get-incident
Get one incident by its id or incident number: title, status, urgency, service, assignees, acknowledgements, pending actions and alert counts.
pagerduty-list-incident-notes
List the notes on one incident, newest first, with their author and time. PagerDuty returns every note in one response and has no pagination for notes, so a very long notes history is cut at 64 KB.
pagerduty-list-incident-log-entries
List one incident's log entries (triggered, notified, acknowledged, assigned, annotated, snoozed, resolved and so on), 25 per page. is_overview true keeps only the most important changes. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-list-incident-alerts
List the alerts grouped into one incident, 10 per page, including each alert's body and details from the monitoring tool that raised it. Incidents opened by hand have no alerts. When more is true, pass offset 10, 20 and so on for the next page.
pagerduty-list-services
List services, by name, 25 per page: name, status, escalation policy and teams. Use it to find the service_id an incident is opened on. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-get-service
Get one service: name, description, status (active, warning, critical, maintenance or disabled), escalation policy, urgency rule and timeouts.
pagerduty-list-oncalls
List who is on call, 25 per page: each entry is a user, the escalation policy and level, the schedule, and the start and end of the shift. Without since/until it shows who is on call now; filter by escalation policy, schedule or user, and earliest true keeps only the next shift per user and level. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-list-escalation-policies
List escalation policies, by name, 25 per page, each with its escalation rules and targets and the services that use it. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-list-schedules
List on-call schedules, by name, 25 per page, with their time zone, users and escalation policies. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-list-users
List users in the account, 25 per page, to find the user id to reassign an incident to: name, email, role, time zone. When more is true, pass offset 25, 50 and so on for the next page.
pagerduty-list-maintenance-windows
List maintenance windows, 25 per page: start and end time, description and the services they cover. filter picks past, future, ongoing, open (ongoing or future) or all. When more is true, pass offset 25, 50 and so on for the next page.
Writes
pagerduty-create-incident
Open one incident on a service, which notifies whoever is on call for it.
fromservice_idtitle
Read back
Read back by incident id; the title must match.
pagerduty-acknowledge-incident
Acknowledge one incident.
fromincident_id
Read back
Read back from the incident; its status must be acknowledged.
pagerduty-resolve-incident
Resolve one incident, which stops its notifications.
fromincident_id
Read back
Read back from the incident; its status must be resolved.
pagerduty-add-note
Add a note to one incident's timeline, visible to everyone who can see the incident (at most 2000 notes per incident).
fromincident_idcontent
Read back
Read back from the incident's notes, newest first; the newest note must be this one, with this text, so a note added by someone else in between makes the read-back fail.
pagerduty-reassign-incident
Reassign one open incident to one user, replacing its current assignees; PagerDuty notifies the new assignee.
fromincident_iduser_id
Read back
Read back from the incident; its first assignee must be that user.
pagerduty-change-incident-urgency
Change one incident's urgency to high or low, which changes how assignees are notified.
fromincident_idurgency
Read back
Read back from the incident; its urgency must match.
pagerduty-snooze-incident
Snooze one acknowledged incident for a fixed time, after which it returns to triggered and notifies again; PagerDuty refuses to snooze an incident that is not acknowledged.
fromincident_idduration
Read back
Read back from the incident; it must still be acknowledged.
pagerduty-merge-incidents
Merge one source incident into an open target incident: the target takes the source's alerts and the source is resolved. Only incidents with alerts, or opened by hand, can be merged, and the target must be open.
fromincident_idsource_incident_id
Read back
Read back from the source incident; it must be resolved with the target as its merge reason. Cannot be undone.
pagerduty-create-maintenance-window
Create a maintenance window on one service: between start_time and end_time PagerDuty opens no incidents on it, so nobody is paged for it.
fromservice_idstart_timeend_timedescription
Read back
Read back by the window's id; its description and service must match.
pagerduty-end-maintenance-window
End one maintenance window: PagerDuty deletes a future window and ends an ongoing one now; a window that has already ended cannot be changed.
maintenance_window_id
Acknowledged
Sent once and acknowledged; not read back: a deleted future window is gone but an ended ongoing window stays readable, so no single read-back fits both.
The brake
Suggested limits
Shared by every agent, session and key in the environment. You set the number.
Suggested approvals
Installed turned off. The reviewer sees the exact request, and the approved request is the one that runs.
Before you rely on it
Use it
npm install -g @anlyonhq/cli
anlyon auth login
# Shows where to create the key, then asks for it without echoing
anlyon connections add pagerduty --secret PAGERDUTY
# From a clone of github.com/AnlyonHQ/anlyon-platform:
# see what would be installed, then switch on one operation
anlyon packs plan ./packs/pagerduty
anlyon packs install ./packs/pagerduty --action create-incident --yesimport { Client } from "@anlyonhq/sdk";
const anlyon = new Client({ apiKey: process.env.ANLYON_AGENT_KEY! });
const { data } = await anlyon.actions.invoke(
"pagerduty-create-incident",
{
"from": "oncall@example.com",
"service_id": "PWIXJZS",
"title": "Checkout error rate above 5%"
},
);
if (data?.pendingApproval) {
console.log("Waiting for a person", data.approvalId);
} else {
console.log(data?.status, data?.responseStatus);
}# Claude Code. Sign in with the browser, no key to paste.
claude mcp add --transport http anlyon https://api.anlyon.com/mcp
# Then ask in plain words. Each operation you switched on is a tool.Prefer to click?
The console does all of it: connect, switch operations on, set the limit, and copy the setup for your agent.
More integrations
API keyPaymentsStripeRefunds under a money ceiling, and every customer, payment and invoice to read.How it works
API keyCodeGitHubIssues, pull requests and code, in the repositories you choose.How it works
API keySupportIntercomSearch the inbox, reply, assign and close, with a person on the reply.How it worksAPI keyCRMAttioPeople, companies, lists, notes and tasks in the CRM startups run on.How it works
API keyDocumentsNotionSearch, read and write pages and databases, only where you shared access.How it works
API keyErrorsSentrySearch issues and events; resolve, assign and archive, with a person on the resolve.How it works
API keyTasksAsanaProjects and tasks: read them, create, update and complete, under a daily limit.How it works
API keyMessagingDiscordRead channels, post, reply and thread as your bot, under a daily limit.How it works
API keyEmailResendSend email from your domain, counted per recipient.How it worksFree tier, no credit card. One command if you use Claude or Cursor.
$ claude mcp add anlyon -- npx -y @anlyonhq/mcp-server