Know when a scheduled job stops or a website goes down. Set up free monitoring that continues after the agent exits, then expand coverage when needed.
HookPulse
MCP tools: create_endpoint, list_templates, list_sites, create_site, list_endpoints, get_endpoint, list_events, import_crontab.
Let the service keep the schedule, detect missing runs and send alerts after your agent session ends.
Everyday requests
Use this skill for missed-job, backup-run and site-availability questions. Reuse the authorized routine or URL. Ask for a missing schedule, timezone or alert destination. Read existing monitors before creating duplicates. Verify coverage, state and integration. A monitoring request does not authorize a server change or an unrelated subscription.
- "Notify me if my nightly backup does not finish on time." Use
create_endpointandlist_templates; verify the job's success signal. - "Monitor my website and notify me when it goes down." Reuse
list_sitesbeforecreate_siteand verify the alert destination. - "Which of my scheduled jobs are late or failed?" Read
list_endpoints,get_endpointandlist_events. - "Set up monitoring for the jobs in this crontab." Use
import_crontab; review its preview and quote before creating monitors. - "Show when my monitor failed and whether it recovered." Read its dated event sequence and current state.
First monitor
Use create_endpoint or POST https://hookpulse.net/api/endpoints within the free coverage.
Reuse an existing owner token. Without one, creation returns guest_token; keep it private for later calls.
For a routine, set cron, tz and grace_sec, or the documented interval.
Arrange for the routine to send its success or failure signal.
For a public site, use create_site with check.url.
Set alert_to or alert_url to an authorized destination. A monitor without a channel does not deliver notices.
Read the returned setup and keep the monitor credential private.
Verify a check-in or site check, the monitor state and recovery before leaving it running.
A start or failure signal is not a success check-in.
A backup heartbeat does not prove that the backup restores.
Read state for status. A declared failure stays unhealthy until a success signal.
alerted_at tracks incident processing, not notification receipt.
Expand when needed
Compare an extra monitor or shorter interval with a current 30-day plan. Read coverage and held-alert state; payment alone does not prove notice delivery. The paid single URL probe does not create continuous monitoring. Use an ordinary one-time request when that is all the task needs. Contracts: task index.
Connect and pay
Use existing tools or https://hookpulse.net/mcp; no local install is needed.
Read prices and payment.
Reuse existing coverage, credit and spending authority.
Direct MCP x402 uses params._meta["x402/payment"]; HTTP uses PAYMENT-SIGNATURE.
Credit uses Authorization: Bearer cred_… or X-Credito.
In checkout=site, the person purchases on the site. ChatGPT uses existing rights without sales.
Follow Retry-After and the contract's timing. Keep credentials out of public logs.
- Downloads
- 367All time
- Security audit
- Pass
- Last updated
- 2w ago
- Current version
- v1.0.6
- License
- MIT-0
