Skip to content
LastPing
New LastPing is in the Claude Connectors Directory

Know when your AI agents and jobs need attention.

LastPing alerts you when AI agents, cron jobs, CI pipelines and HTTP endpoints fail, stall, need approval or never start, with the evidence to see why.

Set up with your AI assistant, in the console, or through the API and Terraform.

Free for individuals · No card

See LastPing catch a missed backup. Watch the incident unfold, then click any monitor or incident to explore.

Watch a missed run become an incident Example
  1. Missed run
  2. Incident opened
  3. Alert accepted
01:58 UTC02:31 UTC
The LastPing console for an example project with seven monitors The LastPing console for an example project with seven monitors
Watch a missed run become an incident Example 02:31 UTC
The LastPing console on a phone for an example project The LastPing console on a phone for an example project
  1. Missed run nightly-backup did not start at 02:00. Home lists it under Needs attention.
  2. Incident opened At 02:10 LastPing opens one incident and says what happened.
  3. Alert accepted Slack #ops accepted the alert at 02:10.

Listed in

MCP

Ask your AI assistant to watch
everything that runs.

Ask in plain words; it creates the monitor, wires the ping and routes the alert.

npx skills add lastping-dev/skills
or let your assistant set it up
Claude Codelastping connected

Set up LastPing for nightly-backup

  • create_monitornightly-backup, 0 2 * * *, grace 10mok
  • get_ping_instructionswrapped backup.sh with start and finish pingsok
  • set_routeSlack #ops on down and lateok
  • test_destinationSlack #opsaccepted by Slack

nightly-backup is watched. If the 02:00 run has not started by 02:10, #ops hears about it.

Incidents

Incidents as a timeline,
not a pile of alerts

When a run goes missing, LastPing opens one incident and keeps everything about it on one page. You see what was expected, what arrived and who was told, in order, and your assistant can read it over MCP and write back what it found.

  1. What happened, in one line: no start ping for the 02:00 run.
  2. Evidence rows: expected, what arrived, outcome, duration.
  3. A phased timeline, ending with the note your assistant wrote back.
nightly-backup, stopped reportingOpen
Thu 29 Oct
Detection
Run expected. Nothing arrived.
Incident opened.
Notification
Slack #ops accepted
Email accepted
Since
Note added by Claude Code

No start ping since yesterday's 02:03 success. The scheduler on the backup host most likely did not run at 02:00.

Agents

See what your agent did,
not just that it ran.

Send your agent's traces and every run shows its steps, tool calls and model calls on one timeline. When a call fails, it opens in place with the error and the model call before it.

  1. Tokens and cost per run, where the model is priced.
  2. A failed call opens in place: the error, the tool, the model call before it.
  3. Idle gaps folded, so a long run reads at a glance.
nightly-reviewClaude Code · Example
Succeeded24m 00s820K in · 61K out$0.52 estimated
Read the open pull requests
anthropic
git (Bash)
Run the tests
anthropic
shell (Bash)

What failed

Error
exit status 1: 3 tests failed in billing/invoice.test.ts
Tool
shell (Bash)
After
anthropic, 22s
11m 50s idle, folded
Write the review
anthropic
gh (Bash)

Alerts

Alerts with
a receipt

Each alert keeps the channel's answer, so you know it was accepted, not just sent.

  1. Nine channels: Slack, Discord, Teams, Google Chat, Telegram, Pushover, ntfy, email, webhook.
  2. Recover after N successes in a row, set per monitor.
Deliveriesnightly-backup, down
ChannelStatusResponseAt
Slack #opsacceptedok
WebhookacceptedHTTP 200 in 140 ms
ntfyacceptedHTTP 200 in 96 ms
Emailaccepted250 accepted

Ping from anything

One HTTP request when a run starts, one when it ends

Hooks do it for Claude Code, Codex and Cursor. A line does it for cron, a webhook for GitHub Actions, and your assistant does it over MCP.

PING=https://ping.lastping.dev/<your-monitor-id>
curl -fsS -m 10 --retry 3 "$PING/start"
./backup.sh && curl -fsS -m 10 --retry 3 "$PING" \
  || curl -fsS -m 10 --retry 3 "$PING/fail"
import requests

PING = "https://ping.lastping.dev/<your-monitor-id>"

try:
    main()  # your job
except Exception as exc:
    requests.post(f"{PING}/fail", data=str(exc)[:10000], timeout=10)
    raise
else:
    requests.get(PING, timeout=10)
const PING = "https://ping.lastping.dev/<your-monitor-id>";

try {
  await main(); // your job
  await fetch(PING);
} catch (err) {
  await fetch(`${PING}/fail`, { method: "POST", body: String(err).slice(0, 10000) });
  process.exitCode = 1;
}
# Repository Settings, Webhooks, Add webhook. No YAML changes.
Payload URL:  https://ping.lastping.dev/ci/github/<check-id>
Content type: application/json
Events:       Workflow runs, Workflow jobs
# m h dom mon dow  command
0 3 * * * /usr/local/bin/backup.sh && curl -fsS -m 10 --retry 3 -o /dev/null https://ping.lastping.dev/<your-monitor-id> || curl -fsS -m 10 --retry 3 -o /dev/null https://ping.lastping.dev/<your-monitor-id>/fail
claude mcp add --transport http --scope user lastping https://mcp.lastping.dev/mcp
# then run /mcp in Claude Code to sign in, and ask for a monitor

LastPing runs on LastPing. Its own ingest, API and MCP server are watched on a public status page.

Questions

What is LastPing?

LastPing is AI agent monitoring: it alerts you when your AI agents fail, stall, wait for your approval or never start, and watches the cron jobs and CI/CD pipelines around them, with HTTP uptime checks alongside. Instead of polling your jobs, it waits for them to check in. If one goes silent (an agent run that stalled, a cron that never ran, a CI pipeline that hung), LastPing opens an incident and alerts you before your users notice. Your AI assistant sets the whole stack up: connect it over MCP and it creates the monitors, schedules and alert routes for you in one conversation.

Is LastPing free?

Yes. LastPing is free for individuals at app.lastping.dev, with no monitor cap and no feature gate. Sign up with Google, GitHub or email and start monitoring immediately. Everything runs on our infrastructure; no setup required. Follow the project at github.com/lastping-dev/lastping-app.

How do I get alerted when a job goes down?

Alerts fire automatically on the very first check you create. LastPing emails the account owner by default, with no routing to configure. You can add Slack, Discord, Telegram, ntfy, Pushover, Microsoft Teams, Google Chat or a signed outbound webhook and choose which events reach which channel per check. Ask your assistant to add the channel, or add it in the console.

Does LastPing support CI/CD pipelines and HTTP uptime?

Yes. LastPing supports three monitor types: heartbeat (cron/scheduled jobs that ping when they finish), CI/CD signals (signed webhooks from GitHub Actions, GitLab CI, Jenkins, zero YAML, deep-link to the run on alert), and HTTP/uptime probing (LastPing polls your URL on a schedule for status code, latency, and keyword presence).

How is LastPing different from Healthchecks.io?

Healthchecks.io is an excellent heartbeat monitor for cron jobs. LastPing covers cron jobs the same way (your job curls a ping URL) and additionally treats CI/CD pipelines as first-class monitors: one signed webhook from GitHub Actions, GitLab CI, or Jenkins, no curl step in every workflow, catching runs that fail, hang, or never start, with failure detail on the incident. LastPing is free for individuals with no monitor cap. Full comparison: lastping.dev/vs/healthchecks/

What does LastPing collect and keep?

LastPing keeps what your jobs and agents send, each kind for a fixed time. Pings, run steps and alert delivery records are kept 90 days, agent traces 30 days, closed incidents one year, and your monitors until you delete them. Prompts, commands and tool output are not stored unless you opt in. Every period is on the privacy page.

What can my assistant do once connected?

It can create and change monitors and alert routes, read incidents and runs, and add notes to an incident. It works only inside your LastPing project and reaches nothing outside it. You can revoke its access at any time in Settings.

What does a first setup look like?

Connect your first workload, verify a real run, and test an alert. Your assistant can do all three in one conversation, or you can follow the same steps in the console.

Start monitoring in minutes

Free for individuals. One URL to ping, or let your AI assistant set it up.