← PingCron blog

Product update

Connect Claude, ChatGPT, and Codex to PingCron

By PingCron · · 5 min read

Your nightly backup misses its check-in. You open the dashboard, copy the status, paste the history into your AI assistant, then go back for the timestamps. Now you have another job: clipboard management.

PingCron now gives your assistant direct, permission-controlled access to monitoring information through the Model Context Protocol (MCP). Connect your own Claude, ChatGPT, or Codex account, approve the access you want to give it, and ask about your scheduled jobs from the conversation you are already in.

Your monitoring context, without the copy and paste.

Included on every PingCron plan, including Free. Start with read-only access using your own assistant account. Custom connector availability depends on your provider, plan, and workspace.

Connect your assistantCreate a free account

Ask about the evidence behind an alert

With the appropriate read permissions, an assistant can list your monitors, inspect their status and retained run history, review recorded incidents and notification attempts, and check your account’s plan and limits. It can also use Cron Doctor to analyze a five-field cron expression with an optional timezone.

Try one of these after connecting:

Use PingCron to list my monitors and their current status. Point out anything that needs attention.
Review the retained incidents for my nightly backup. Separate recorded evidence from possible causes, and tell me what logs are still missing.
Use PingCron to analyze 0 2 * * * in America/Los_Angeles. Explain the schedule and any daylight-saving caveats.

A missing heartbeat does not reveal why a job failed. PingCron’s diagnosis tools return available monitoring evidence and identify what is missing; they do not invent a root cause. Your scheduler logs, process exit details, and output checks still matter. Cron analysis does not execute your commands.

Connect your assistant

  1. Sign in to PingCron and open AI Agents.
  2. In your assistant’s remote MCP or custom connector setup, use this server URL:
https://api.pingcron.io/mcp
  1. Choose OAuth when asked for authentication, and dynamic client registration (DCR) if a registration method is offered. PingCron does not require a static client ID or client secret for this flow. Complete the PingCron sign-in and review the client, return address, and requested permissions. Start with read-only access.
  2. Enable the connection in a conversation and ask the assistant to list your monitors. If your client needs a fresh session to load new tools, start one.

Use Claude’s custom connector flow or ChatGPT’s custom MCP app setup where available. These are custom connections; PingCron does not need to appear in a provider’s public app directory. Availability and setup controls depend on your client, plan, and workspace policy. See the current Claude connector instructions and ChatGPT custom MCP guide.

For Codex CLI, add the server and start OAuth with read permissions:

codex mcp add pingcron --url https://api.pingcron.io/mcp
codex mcp login pingcron --oauth-client-registration dcr --scopes account:read,monitors:read,history:read,incidents:read,diagnostics:read,alerts:read

The add command may begin sign-in automatically. Options vary by Codex version; check codex mcp login --help if your installation differs. You do not need to paste an OpenAI or Anthropic API key into PingCron.

Provider setup and compatibility

Codex

Existing OAuth connection verified on October 9, 2026

Use a Codex client with remote MCP and OAuth enabled. Organization policy may restrict custom servers.

  1. Add a remote MCP server using the URL below. In Codex CLI, use codex mcp add pingcron --url followed by that URL.
  2. Start OAuth sign-in. Select dynamic client registration (DCR) if offered. Review the PingCron account and read permissions.
  3. Start a fresh conversation if tools do not appear. Request the report below.
https://api.pingcron.io/mcp

Account, monitor and cron-analysis calls were verified with native OAuth. The new missed-check-in report awaits live Codex verification.

Provider documentation checked October 10, 2026. Availability can change.

Verify a missed-check-in report

The report requires incidents:read. Monitor names additionally require monitors:read. Creation and editing require separate permissions. This report is under review and may be unavailable until your backend is updated.

Use PingCron to answer: Which scheduled jobs missed their check-ins in the last 24 hours? Use get_missed_checkins with my timezone (ask me if unknown). State the exact returned window and timezone. Follow every next_after with the same from, to and timezone, even if a page has no matches. Identify monitor IDs and names when permitted, recorded deadlines and event timestamps, and recovery as of the window end. Report unclassified events, retention limits, and any incomplete traversal or failed calls. Do not infer root causes or job-output correctness. If the tool is unavailable, say so instead of treating a generic failure list as an exact report.

Review grants and Recent activity in AI Agents. Revoke the grant there and disconnect the provider plugin. An admitted read may finish within five seconds. Reconnect expired authorization through OAuth; do not paste credentials into chat. For rate limits, wait before a bounded retry. For unavailable tools or partial data, report the limitation.

Connecting grants access; it does not start continuous observation. Any provider routine needs separate instructions and scheduling. PingCron monitors check-ins. It does not execute, retry or repair jobs, determine root causes, or verify business-output correctness.

You choose what the agent can do

Read access and write access are separate. If you explicitly grant create or update permission, an agent can create a monitor or edit its name, schedule, and grace period. Writes have retry protection, and updates check that the monitor has not changed since the agent read it.

Creating a monitor does not install a heartbeat in your job. Get the private heartbeat URL from the dashboard, add it after the required work succeeds, and verify the first receipt and an alert test. The agent interface does not expose heartbeat URLs, private alert destinations, billing changes, monitor deletion, or commands that run your infrastructure.

Review credentials, grants, and recent activity in AI Agents. Revoke a grant there to prevent future requests; an already admitted read may finish. Connected monitoring data is shared with the assistant you authorize, so use an account and provider policy appropriate for that data.

Your agent account, your PingCron limits

AI assistant connections are included on every PingCron plan, including Free. Keep using your own assistant account. PingCron serves monitoring tools; this integration does not make paid OpenAI or Anthropic model API calls on your behalf. Your provider’s subscription, usage limits, and any applicable charges still apply. Your PingCron plan continues to determine monitor limits and available history.

We verified connections with Codex, Claude, and ChatGPT on isolated staging. All three completed OAuth and account, monitor-list, and cron-analysis calls. That is a tested starting point, not a promise that every tool works in every client version. The integration is now live on PingCron.

Start with one useful question

Connect an assistant with read-only access and ask it which monitors need attention. Keep the answer tied to recorded evidence, then choose the next action yourself.

Open AI Agents

New to PingCron? Create an account. Building your own integration? Read the AI product reference.