My AI agents sent 51 cold outreach messages while I slept. Here's the $200 Mac Mini setup that made it happen.

Last night while I was sleeping, my AI agents:

• Identified 327 insurance prospects and sent 51 personalized outreach messages
• Checked my financial data API revenue (up to $0.60 from 3 paying customers)
• Generated and published a 10-tweet macro market intelligence thread to X
• Left 3 substantive comments on Reddit in my niche
• Ran a heartbeat check on all systems and filed a status report

I didn't touch a keyboard. I didn't check a single notification. I just... slept.

This isn't a pitch. This is my actual Monday morning.

The Problem Nobody Talks About

Here's the thing about being a solopreneur: the business isn't the hard part. The admin around the business is the hard part.

Studies consistently show solopreneurs spend 60-70% of their time on operational overhead — lead follow-up, data checking, social posting, report generation, inbox triage. Not building. Not selling. Just... maintaining.

I run an insurance operation and a financial data API called Mercury402 (serves deterministic financial data via x402 micropayments — think Stripe for API calls, but with crypto-native payments). Both need daily attention. Both have repetitive tasks that compound if neglected.

I don't have employees. I don't want employees. I want leverage.

So I built agents instead.

The Exact Architecture

Hardware: A used Mac Mini sitting on my desk. That's it. No cloud servers, no Kubernetes clusters, no DevOps team.

Software: OpenClaw — an open-source agent framework that connects AI models to tools (browser, file system, APIs, Telegram, cron jobs). 331K+ GitHub stars. The agent equivalent of what WordPress was for websites.

Agents:

Forge: Strategy + Execution — File ops, API calls, cron jobs, research, memory management, coordination

Phoenix: Browser + Social — X/Twitter posting, Reddit engagement, web scraping, visual tasks

Control layer: Telegram. I send commands like "run the weekly macro report" or "check Mercury402 revenue" from my phone. Agents execute, report back with results + next action.

Cost: Under $50/month in API calls. The Mac Mini was a one-time purchase. OpenClaw is free.

What Runs on Autopilot

Here's the actual daily/weekly schedule:

Every 6 hours (9am–10pm):

• Revenue monitor hits Mercury402's API, pulls call counts, customer count, top endpoints
• Compares against milestone thresholds ($0.50, $1, $5, $10...)
• If a milestone is crossed → auto-posts to X with the milestone announcement

Daily:

• Heartbeat check: are all services alive? Gateway running? Browser sessions stable?
• Engagement rounds: Moltbook comments (niche community), Reddit replies in relevant subs
• Research missions: overnight deep-dives on competitive intel, market data

Weekly (Monday 7am):

• Macro intelligence report: pulls yield curve, CPI, payrolls, GDP, FRED indicators
• Synthesizes into a Goldman-Sachs-style briefing (~800 words)
• Auto-publishes as a 10-tweet thread to X

Overnight research missions:

• 20+ subagents spawn in parallel, each tackling a specific research question
• Last run produced 16 research files (4,500+ lines of analysis), a 4-page website redesign, and a morning briefing with prioritized action items
• I wake up to a briefing, not a to-do list

The Real Numbers (No Hype)

I want to be honest here because IH audiences can smell fluff:

• Mercury402 revenue: $0.60 total from 3 customers. Not life-changing. But it's verified on-chain revenue from strangers on the internet that my agents monitor and celebrate automatically. The system works.
• Insurance prospects: 327 identified through automated research. 51 personalized messages sent in a single overnight session. That's a week of manual work done before breakfast.
• Overnight research output: Equivalent to hiring a junior analyst for 20 hours. Cost: ~$8 in API calls.

The Insight That Changed Everything

Most people approach AI agents wrong. They try to build a standalone AI business — "my agent makes money on its own."

That's the hard path. The easy path is different:

Agents are leverage layers on existing businesses, not standalone businesses.

I didn't build an agent to replace my insurance business. I built an agent to handle the 60% of my insurance work that's repetitive. I didn't build an agent to create a new API. I built an agent to monitor, promote, and optimize an API I already built.

The agent doesn't replace the business. It makes the business scale without me being present.

Think of it like this: if you're a solopreneur doing $5K/month and spending 60% of your time on admin, you're effectively paying yourself $25/hr for operational work. An agent that handles that admin for $50/month is a 50x ROI on the spot.

What's Next

I'm working on productizing this as a "Solopreneur Operations Agent" — essentially a pre-configured OpenClaw setup designed for solopreneurs who run 2-3 revenue streams and need someone (something?) to handle the operational overhead.

The vision: you tell it what businesses you run, it configures the monitoring, outreach, reporting, and engagement automatically. You wake up to briefings, not backlogs.

The Setup I'd Recommend Starting With

If you're curious about doing this yourself:

1. Get OpenClaw running — it's free, open-source, runs on any machine
2. Connect Telegram — this becomes your command center
3. Start with ONE repetitive task — revenue checking, lead monitoring, social posting
4. Add cron jobs — let it run on schedule, report results
5. Add a second agent when you need browser automation — social posting, web scraping, research

Don't try to automate everything at once. Start with the task you hate most. Let the agent earn its keep. Then expand.

Building in public as @Mercuryclaw1 on X. Mercury402 is live at mercury402.uk. OpenClaw is at github.com/openclaw/openclaw.