OpenClaw Heartbeats: How Smart Businesses Use Proactive AI Monitoring

Most people set up OpenClaw, connect it to Telegram or Discord, and then wait for something to happen. They type a command, get a response, type another. That’s fine. But without OpenClaw heartbeats configured, you’re leaving about 80% of the value on the table.

OpenClaw heartbeats change the dynamic completely. Instead of your agent waiting passively for instructions, heartbeats make it proactive. Every 30 minutes (or whatever interval you configure), your agent wakes up, checks a short list of things that matter to your business, and either handles them silently or sends you a quick alert.

Think of it like the difference between checking your security cameras once a day versus having a guard who patrols every half hour. The cameras are useful. The guard is better.

This article breaks down how heartbeats actually work, what different types of businesses are doing with them, and the configuration details you need to get started.

How OpenClaw heartbeats work (the technical version, simplified)

A heartbeat is a scheduled agent turn. At a set interval, OpenClaw sends your agent a prompt (either the default or a custom one you write). The agent reads a file called HEARTBEAT.md in your workspace, runs through whatever checklist you put there, and decides one of two things:

1. Nothing needs attention. It replies HEARTBEAT_OK, and you never see a message. Quiet. Clean.
2. Something needs attention. It sends you an alert with context about what changed and, if configured to do so, takes action on its own.

The default interval is 30 minutes. You can set it to 5 minutes if you’re in high-velocity mode, or stretch it to 2 hours if you just want a casual background watch. You can also restrict heartbeats to active hours so your phone isn’t buzzing at 3am with “all clear” messages.

Here’s the basic config in your OpenClaw gateway:

{
  agents: {
    defaults: {
      heartbeat: {
        every: "30m",
        target: "last",
        activeHours: {
          start: "08:00",
          end: "22:00"
        }
      }
    }
  }
}

And here’s a sample HEARTBEAT.md checklist:

# Heartbeat checklist
- Check email for urgent messages
- Review calendar for events in next 2 hours
- If a background task finished, summarize results
- If idle for 8+ hours, send a brief check-in

The beauty of this system is that HEARTBEAT.md is just a plain text file. No coding required. You write what you want checked in plain English, and the agent follows it.

Why heartbeats matter more than most people realize

There’s a subtle but important distinction between heartbeats and cron jobs. Cron jobs fire at exact times and run in isolated sessions. They’re great for “send this report every Monday at 9am.” Heartbeats run in the main session, which means they share conversational context.

That context awareness is what makes heartbeats so powerful. Your agent knows what you’ve been working on. It knows you had a meeting at 2pm. It knows you asked about a client proposal yesterday. So when the heartbeat fires and scans your inbox, it can connect the dots: “Hey, the client you were discussing yesterday just replied to your proposal email.”

A cron job can’t do that. It runs blind.

The other big advantage is cost efficiency. Instead of running five separate cron jobs to check email, calendar, notifications, project status, and weather, a single heartbeat batches all of those into one agent turn. One API call instead of five. According to the OpenClaw documentation, this is actually the recommended pattern for periodic monitoring.

OpenClaw heartbeats for e-commerce businesses

E-commerce runs on urgency. A customer complaint sitting unanswered for 4 hours can become a negative review. A supplier email about a shipping delay needs an immediate response before you oversell inventory.

An e-commerce heartbeat checklist might look like this:

# E-commerce heartbeat
- Check support inbox for unanswered tickets older than 1 hour
- Scan email for supplier shipping updates
- Review Shopify notifications for order issues or chargebacks
- Check inventory alerts for items below reorder threshold
- If any customer rated us 1-2 stars, flag immediately

The agent checks every 15-30 minutes during business hours. Most of the time, it returns HEARTBEAT_OK. But when a supplier emails about a 3-day shipping delay on your best-selling product, you know about it within the hour instead of discovering it when customers start complaining.

For stores doing $50,000+ per month in revenue, even catching one stockout early or responding to one angry customer faster can be worth hundreds of dollars. The heartbeat itself costs almost nothing because quiet checks (HEARTBEAT_OK) are cheap and fast.

OpenClaw heartbeats for consultants and agencies

Consultants live and die by responsiveness. Missing a hot lead’s email by half a day can mean losing a $10,000 contract to someone who replied faster.

A consulting heartbeat setup:

# Consulting heartbeat
- Check email for new inquiries or RFP responses
- Review calendar for meetings in next 2 hours and prep reminders
- Check project management tool for overdue tasks
- Scan Slack/Teams for direct messages from clients
- If any proposal is pending response for 48+ hours, remind me

The context awareness really shines here. When your agent processes a heartbeat and sees a new email from a prospect, it also knows you’re booked solid today because it can see your calendar. So instead of just saying “new email from John,” it might say “new email from John, but you’re in meetings until 4pm. Want me to send a quick acknowledgment?”

That’s the difference between a dumb alert and an intelligent one.

OpenClaw heartbeats for real estate professionals

Real estate has a unique timing problem. New listings hit the market at unpredictable times. A buyer lead from Zillow or Realtor.com goes cold in hours, not days. And showing schedules change constantly.

A real estate heartbeat:

# Real estate heartbeat
- Check email for new buyer/seller inquiries
- Scan MLS alerts for new listings matching active buyer criteria
- Review showing schedule for conflicts or cancellations
- Check for new reviews on Google Business Profile
- If a contract deadline is within 48 hours, send reminder

The 30-minute default interval works well here. Fast enough to catch a new hot lead within the hour. Slow enough that you’re not getting constant buzzes while showing a property.

And because heartbeats run in the main session, the agent remembers your buyer’s criteria from previous conversations. When a new MLS listing matches, it doesn’t just say “new listing in 90210.” It can say “this 3-bed in 90210 matches what Sarah Chen has been looking for, and it’s $15K under her budget.”

OpenClaw heartbeats for content creators and media businesses

Content businesses need to monitor multiple channels, track competitor activity, and catch trending topics before they’re old news.

# Content business heartbeat
- Check social media mentions and DMs
- Scan RSS feeds for breaking news in our niche
- Review comment sections for unanswered questions
- Check analytics for traffic spikes (could indicate viral content)
- If a collaboration email came in, flag it

A content creator running a newsletter might set heartbeats to every 60 minutes during active hours. The agent quietly watches RSS feeds, social mentions, and email. When something relevant breaks, you get a single, contextualized alert instead of drowning in raw notifications from six different platforms.

The cost question: how much do OpenClaw heartbeats actually cost to run?

This is where heartbeats get interesting from a business perspective. A community member wrote about this pattern in a detailed breakdown on DEV Community: run cheap, deterministic checks first, and only involve the AI model when there’s actual signal to process.

Here’s the math. At 30-minute intervals during a 14-hour active window (8am to 10pm), you get 28 heartbeats per day. Most of them return HEARTBEAT_OK almost immediately because nothing changed. The model reads your checklist, checks the relevant sources, finds nothing urgent, and exits. That’s maybe 500-1,000 tokens per quiet heartbeat.

At current API pricing, 28 quiet heartbeats per day using Claude Sonnet costs roughly $0.10-0.20 per day. Maybe $3-6 per month. When a heartbeat does find something important, it uses more tokens for the alert and any follow-up actions, but those events are the minority.

Compare that to the cost of missing an urgent email for 4 hours, or forgetting about a contract deadline, or being the last to know about a supply chain issue. The ROI math isn’t close.

You can push costs even lower by using a smaller model for heartbeat checks. The config supports per-agent model overrides:

{
  agents: {
    defaults: {
      heartbeat: {
        every: "30m",
        model: "anthropic/claude-haiku",
        target: "last"
      }
    }
  }
}

Setting up heartbeats: the practical steps

Getting heartbeats running requires three things:

1. Enable heartbeats in your OpenClaw config. The default interval is 30 minutes. Set target: "last" so alerts go to your last active channel (Telegram, Discord, etc.). Add activeHours so you’re not getting pinged at 2am.

2. Write your HEARTBEAT.md file. This goes in your agent’s workspace root. Write it in plain English. Be specific about what you want checked and how urgent each item is. Vague instructions like “keep an eye on things” produce vague results.

3. Connect the services your agent needs to check. If you want email monitoring, your agent needs access to your email via the gog CLI or similar integration. Calendar checks need calendar API access. Each service connection adds capability to your heartbeat.

The tricky part isn’t any single step. It’s getting the combination right. A heartbeat checklist that’s too aggressive wastes tokens. One that’s too conservative misses important things. The activeHours window needs to match your actual work schedule. And service connections need proper authentication that doesn’t expire every 48 hours.

This is where most DIY setups struggle. The heartbeat feature itself is straightforward. But tuning it to actually match how your business operates takes iteration and experience with the platform.

Heartbeats vs. cron jobs: when to use which

A quick reference, because this trips people up:

Use heartbeats for: inbox monitoring, calendar awareness, notification batching, background project checks, anything where context matters and timing can flex by a few minutes.

Use cron jobs for: exact-time delivery (Monday 9am report), standalone heavy tasks (weekly analysis), one-shot reminders, anything that should run in isolation without main session context.

Use both together for: a morning cron that generates a daily brief at 7am, plus heartbeats every 30 minutes that catch urgent items throughout the day. This is the setup most businesses end up with, and it works well because each mechanism handles what it’s best at.

Common mistakes with OpenClaw heartbeats

A few patterns that consistently cause problems:

Setting the interval too short. A 5-minute heartbeat sounds great in theory. In practice, you burn tokens on 288 checks per day instead of 28, and the improvement in response time is marginal for most businesses. Start at 30 minutes. Drop to 15 only if you have a documented need for faster monitoring.

Making the checklist too long. If your HEARTBEAT.md has 15 items, each heartbeat takes longer and costs more. Focus on 4-6 high-value checks. Move everything else to dedicated cron jobs.

Forgetting active hours. Without activeHours configured, your agent runs heartbeats at 3am. Even though quiet heartbeats don’t send messages, they still use API tokens. Set a reasonable window.

Not testing the HEARTBEAT_OK flow. After setting up heartbeats, watch the logs for a day. Make sure quiet heartbeats are actually returning HEARTBEAT_OK and not sending you unnecessary messages. A noisy heartbeat defeats the purpose.