Always-on agents · the series

Always-on agents, one post a day

I run a two-person company on agents that work while I sleep. This is where I explain what always-on agents are, who is shipping them, and how I would set one up on your own machine for nothing. One new post every morning from October 20, 2026 to November 21, 2026.

In short

An always-on agent is a program that keeps a goal, watches a clock or a stream of events, and acts without being asked each time. In 2026 the big labs started selling them as products. You can also run one yourself for free, on almost any computer. This series covers all three: what they are, who makes them, and how I set mine up.

Understand

Start here if you have only seen the headlines. I define the term the way I use it, keep a running list of who ships what, and say what it all costs.

20 OctWhat an always-on agent is, and what it is not21 OctEvery company shipping always-on agents in 202622 OctWhat always-on agents cost: dots, Muse, Spark and the do-it-yourself route23 OctAlways-on agent, chatbot or workflow automation: which one you actually need24 OctBefore you give an agent your inbox: permissions, identities and what researchers found in OpenClaw

Same use cases, free tools

Each of these rebuilds one launch demo from a big vendor with free software, so you can judge the real thing against something you own.

29 OctBuild a Slack bug-triage agent like the OpenAI dots demo, with free software30 OctA morning brief agent: inbox, calendar, numbers and news at seven, every day2 NovAn inbox and calendar agent that works while your laptop is closed7 NovA meeting-prep agent that briefs you before every call10 NovA competitor and brand watch agent that reads the feeds every hour13 NovA prospecting agent that watches for buying signals16 NovAn always-on coding agent that opens pull requests overnight19 NovAn always-on social publishing agent: build it, or use Leo

Set it up anywhere

Twelve guides that share one ten-step spine. Each page covers only what is different on that system, and I tested the commands on a fresh install before publishing.

25 OctHow to run an always-on agent in the cloud: free tiers, a cheap VPS or a hosted runtime26 OctHow to keep an agent running 24/7 on a Mac27 OctHow to run an always-on agent on Windows with Task Scheduler and WSL28 OctHow to run an always-on agent on Linux: the core steps31 OctRunning an always-on agent on Ubuntu3 NovRunning an always-on agent on Debian6 NovRunning an always-on agent on Fedora9 NovRunning an always-on agent on Arch Linux12 NovRunning an always-on agent on a Raspberry Pi15 NovRunning an always-on agent on openSUSE18 NovRunning an always-on agent on NixOS20 NovRunning an always-on agent on Alpine Linux

Open source and free

The projects I would actually run, the free models that can hold a tool loop together, and the workflow tools that were always-on before the word existed.

1 NovOpen-source always-on agents compared: OpenClaw, Hermes Agent and the rest4 NovOpenClaw from zero: gateway, Telegram, scheduled jobs and the security checklist8 NovHermes Agent on a five-dollar server11 NovRun an always-on agent with no API bill: Ollama, small models and their limits17 Novn8n, Huginn and Node-RED: always-on behaviour without an agent framework

In real life

What people run agents for, where they fit in a marketing team, and the honest account of how my own company runs on them.

5 NovTwenty-five things people actually run always-on agents for14 NovAlways-on agents for marketing teams: listening, publishing, reporting21 NovHow a two-person company runs on always-on agents

Who is shipping what

The short version of the rolling headlines post, which goes up on October 21, 2026, which I update as announcements land. Dates are the primary announcement, not the press cycle.

CompanyProductWhat it does when nobody is typingSince
OpenAIdotsTakes on an ongoing job from inside ChatGPT, on its own cloud computer29 Sep 2026
MetaMuseA personal agent for long-running tasks, on a dedicated virtual machine8 Sep 2026
GoogleGemini SparkKeeps working on Google Cloud after your laptop closes19 May 2026
MicrosoftAutopilot (was Scout)Acts across Microsoft 365 under its own identity2 Jun 2026
AnthropicManaged Agents, Claude Code scheduled tasksAgents that run on a schedule or on an event, hostedApr to Jun 2026
SalesforceAgentforcePursues sales goals across days and weeks with memory11 Sep 2026
WorkdayAutopilot AgentsCarries out HR, finance and IT work, not just answers13 Oct 2026
ServiceNowProject ArcA desktop agent that executes across enterprise tools5 May 2026
OpenClawOpenClawOpen source, MIT; the self-hosted favourite, often on a Mac mini2026
Nous ResearchHermes AgentOpen source, MIT, learns skills from its own past workFeb 2026

Words you will see

Always-on agent
A program with a goal that keeps running, watches for triggers and acts without a fresh prompt. Persistence is the defining property, not literal 24/7 compute.
Background agent
The same idea, named for where it runs: out of sight, while you do something else.
dot
OpenAI’s name for one always-on agent inside ChatGPT.
Scheduled task
Anthropic’s name for a saved Claude Code job that runs in its cloud on a schedule, an API call or a GitHub event. It was called a routine until recently, and the web address still says so.
Gateway
The long-lived process in OpenClaw and Hermes Agent that holds the connection to your chat apps and runs scheduled jobs.
Heartbeat
A periodic check-in the agent sends so you know it is alive. Silence is the alarm.
cron
The classic Unix scheduler: run this command at these times.
Daemon
A program that runs in the background with no window, started by the system.
systemd
The service manager on most Linux systems. It starts your agent at boot and restarts it when it dies.
launchd
The macOS equivalent of systemd. Our publishing agents run under it.
Task Scheduler
The Windows equivalent, usually used to start WSL, which then runs systemd.
MCP
Model Context Protocol, the open standard that lets an agent reach tools and data sources.
Local model
A language model running on your own hardware, through a tool like Ollama, so there is no API bill.
Lera Leonteva
Lera Leonteva
CEO & Founder, Leo AI

Lera Leonteva is the founder and CEO of Leo AI, where she works on turning long-form footage into short-form video that earns attention on its own merits. She writes Video Scaling Strategies, taking one piece of content that worked and one that did not, and using what the market is doing to explain the gap.

The always-on agent I run for social video

Leo watches what your audience is responding to, picks a clip from footage you already have, edits it and books it on the calendar. Every hour, without being asked. It is the agent this whole series grew out of.

Keep going with the series

Every post in the Always-on agents series is on one page, newest first. If you would rather see the agent I run for social video, here it is.