Updated 2026-09-29

Server alerts in Telegram: investigate when you are away from your desk

A container starts restarting while you are away from your computer. A useful alert should help you decide what to inspect next, not just tell you something is red. OpsMate can send patrol alerts to your linked Telegram private chat and offer a route into AI-assisted investigation.

This guide covers receiving an alert and following up. An alert being sent does not mean someone has read it, and a suggested command does not mean the problem has been fixed.

First, link your account

1. Open the Telegram binding settings in the OpsMate console and follow the displayed binding flow.
2. Use the bot or deep link shown by the console. Complete the identity confirmation in your private conversation with the bot.
3. Check that binding succeeded and that the relevant patrol and notification settings are configured.

Server operations belong in the bot's private chat, not a public community group. Joining a group is not the same as binding your account.

Read the alert before taking action

An L3 patrol recommendation card can show the server, check, symptom summary, severity, suggested checks and problem status. The card offers “📋 View and handle” and “🔗 View on web”. Alert cards arrive in English or Chinese: your OpsMate account language if you've set one, otherwise the Telegram language last seen in that chat, otherwise English. You can switch with /language in the bot or set it on the web.

L3 is a recommendation requiring human review, not an automatic production repair. Start by checking the affected service and time window. Severity helps prioritize attention; it does not establish the root cause.

Example: a container keeps restarting

The following uses an illustrative container name, not a customer system:

docker logs --tail 200 demo-api-1
docker inspect -f '{{.State.ExitCode}} {{.State.Error}}' demo-api-1

Review the startup error and exit information before deciding whether to investigate a dependency, configuration or memory pressure. Repeated restarts alone do not prove an out-of-memory failure. Replace the example name only after confirming the correct server and container.

Continue investigating in the private chat

The card's “📋 View and handle” button opens the problem details: the issue, severity, server, status, evidence and when it was last updated, with the buttons “Recollect”, “Snooze”, “Intervene”, “AI advice” and “Mark resolved”. “AI advice” gives text suggestions only and runs no commands. “Intervene” is for describing the situation, and you finish it with /done or /cancel. Neither connects to your server.

To have the AI run checks on the server, enter AI SSH mode: in the private chat, send /servers, pick the server and tap “🤖 AI chat”. Then describe a bounded request:

Check recent errors for the affected service without changing anything. Summarize the evidence and possible causes. Do not restart services or change configuration; list proposed changes for me to review first.

The AI chooses the commands, and each one goes through a safety check first. Commands that pass run directly and their output appears as “📟 Terminal output”; commands that don't pass wait as “⚠️ Command pending confirmation” until you tap “✅ Approve”. AI SSH mode is mainly for troubleshooting: dangerous commands are blocked. For what OpsMate does and doesn't do on its own, see the FAQ.

Review the actual commands and results. Asking the AI not to change anything is not a technical permission boundary. Keep SSH permissions appropriate and verify any proposed change before approving it. Use the web entry when you need the full problem details or a larger terminal workspace.

Remote SSH needs cloud-accessible credentials

Receiving alerts in Telegram does not need any cloud credentials. Only remote SSH from Telegram (AI SSH mode) does: for the Telegram service to connect to a server, that server needs an SSH key stored in the OpsMate cloud. Without one, the bot tells you that no SSH key is configured and AI SSH mode can't be used. Credentials stored only in the desktop app do not become available to Telegram merely because you link your account.

Do not upload production credentials just to try the mobile workflow. Choose the credential arrangement that fits your security policy. Local credential storage also does not mean AI processing is local: in AI SSH mode, command output is sent to cloud AI for analysis, and in the web app it is sent after you click Analyze. Remove secrets and personal information before sharing logs or screenshots.

Once remote SSH is set up, SSH from your phone is painful: troubleshoot your server from Telegram instead walks through AI SSH mode: which commands run directly, which wait for your approval, and when to go back to a computer.

If no alert arrives

Check account binding, the private bot conversation, patrol execution and notification settings. A recorded anomaly and a delivered notification are different events. If the bot reports a permission or connection error, verify the selected account and server rather than pasting SSH credentials into the chat.

Try a small, reviewable workflow

Start with a test server and an existing patrol notification. Read the evidence, request a limited follow-up check, and return to the web workspace for any change that needs closer review. For the desktop workflow, see AI and SSH commands in one workspace.

Troubleshooting guides