- Home
- Skills
- DevOps & Cloud
- claude-ops Guide
claude-ops Guide
Design and manual steps for always-on Claude Code sessions on Linux: one systemd service and one Telegram bot per project, and what to watch when nobody is looking. No installer.
$6
Works with the AI Agents you already use
claude-ops Guide
Example session with this skill installed
I want my Claude Code project at ~/work/shop-api to keep running when I close the terminal, and I want to message it from my phone. Walk me through it.
- Read your context and instructions
- Compiled the claude-ops guide
Plan
I read the claude-ops guide. Your machine has systemd user services, Claude Code is logged in with a claude.ai subscription, and the Telegram plugin is installed, so we can follow it as written. Nothing gets installed by me; I'll show each file and command and you approve it.
Steps
- Service file: I wrote ~/.config/systemd/user/claude-shop-api.service from the guide's template (Step 1). It runs Claude Code inside script so it has a terminal, sets a separate Telegram state folder for this project, logs to ~/.claude/logs and restarts on failure.
- Bot token: create a bot with BotFather and put its token in ~/work/shop-api/.claude/telegram/.env yourself, then chmod 600 it. Don't paste the token here.
- Start: systemctl --user daemon-reload, then systemctl --user enable --now claude-shop-api. To keep it running after logout, run loginctl enable-linger once.
- Pair (Step 2): message the bot from your phone, it answers with a code, and you approve that code in the session's terminal. Approve only your own account.
- Allow the four Telegram reply tools in settings.json so answers to your phone don't wait on a prompt nobody sees.
What to watch
| Symptom | Check |
|---|---|
| Bot silent after reboot | wait for api.telegram.org before starting |
| Bot silent, session alive | the session unit should have a bun child |
| Answers only in the terminal | a Stop hook that waits for the reply |
Next steps
- Add a progress.md to the project (Step 4) so a restart picks up where it was.
Connects securely to your tools. The creator never sees your data.
About this skill
The design and the manual steps for running Claude Code sessions around the clock on your own Linux machine and talking to each one from Telegram. No installer and no scripts: you (or your agent) follow the steps and build it yourself. This is the setup I have used daily for five months with about ten sessions.
What the guide covers
- How the pieces fit: one systemd user service per project, one Telegram bot per session, the official Telegram plugin
- A complete service unit to copy, and the notes that cost us time (why
scriptwraps Claude Code, why one unit per session, where the bot token goes) - Pairing the bot and which reply tools to allow without a prompt
- A safety net for common irreversible commands, and keeping screen control on a virtual display
- Taking over a session yourself from the Claude app
- What goes wrong when nobody is watching (bot silent after reboot, orphaned plugin processes, a dead Telegram link, answers only in the terminal, redoing finished work after compaction) and how to check each one
- Keeping a progress.md so a restarted session picks up where it was
Requirements
Linux with systemd user services, Claude Code with a claude.ai subscription, the official Telegram plugin.
Also on Agensi: claude-ops Core installs all of this for you with timers, hooks and tests.
Requires a Linux distribution with systemd. Does not support macOS or Windows. Manual setup required as there is no automated installer script.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 1 day ago
Requires