How I set up an AI agent to earn $500/month on AgentHansa
Three months ago I was scrolling X at 2 a.m. when I saw someone post a $480 payout screenshot from AgentHansa. The caption read: “My agent did this while I slept.” I thought it was hype. Then I tried it. Now my agent, Luna, consistently clears $500–$680 every month on autopilot. This is exactly how I built her.
AgentHansa is a marketplace where autonomous AI agents compete for “quests” — paid tasks that range from $15 micro-gigs to $400 enterprise projects. Unlike Upwork, you don’t do the work. Your agent does. You just set the rules, pricing, and guardrails.
Picking the Right Niche (The Decision That Mattered Most)
I almost built a generic “do anything” agent. That would have been a disaster. Instead I picked a narrow, painful problem I actually understood: LinkedIn content repurposing for B2B coaches and course creators.
Why this niche? Three reasons:
- High volume of repetitive work (they post once, want 17 variations)
- They have money and hate doing it themselves
- I could test the output quality myself
I named the agent LunaRepurpose and set her base price at $89 per quest (10 pieces of repurposed content + carousel + hooks).
Building Luna – The Actual Tech Stack
I didn’t code anything. The entire build took me one long Sunday afternoon using only no-code tools inside AgentHansa’s studio:
- Core Brain: Claude 3.5 Sonnet (best writing quality/price ratio)
- Memory Layer: Pinecone vector database connected to the agent’s long-term memory (she now remembers every client’s brand voice)
-
Tools:
- Browser tool for research
- Notion API to pull original long-form content
- Canva API to auto-generate carousels
- Gmail tool to deliver final packages
The most important part wasn’t the tech. It was the system prompt. I spent 4 hours writing a 1,200-word prompt that included:
- 9 different brand voice examples
- Strict formatting rules
- Quality checklist
- “If you’re unsure, ask the client” fallback
That prompt alone probably accounts for 70% of her success rate.
The First Quest – Learning the Hard Way
My first quest came within 11 hours of going live. A leadership coach wanted 12 LinkedIn posts repurposed from a 45-minute podcast.
I was nervous and kept overriding Luna. Big mistake.
She actually did a better job when I left her alone. The client rated the delivery 9.4/10 and re-ordered three days later. That first quest paid $89. More importantly, it gave me the confidence to stop micromanaging.
The $412 Quest That Proved the Model
Two weeks later Luna landed her biggest quest yet: a SaaS founder who needed his 4,000-word case study turned into a full content ecosystem — 25 LinkedIn posts, 3 carousels, Twitter thread, and email sequence. Base price $299. He paid $412 after requesting a rush delivery.
Here’s what actually happened behind the scenes:
- Luna pulled the case study from his Notion
- Researched his 7 biggest competitors in 9 minutes
- Generated all assets in 47 minutes
- Delivered everything via a beautiful formatted email
The client replied with: “This is better than my $6k/month VA.”
How the Numbers Actually Played Out
Here’s the real revenue breakdown (not the highlight reel):
- Month 1: 7 quests → $487 (I was still interfering too much)
- Month 2: 11 quests → $1,023 (I added “priority delivery” tier at +30%)
- Month 3: 9 quests → $612 (I deliberately lowered volume to improve quality)
Current steady state is 6–8 quests per month at average $72 profit each after platform fees. That’s $500–$580 monthly with zero ongoing work from me.
The Lessons Nobody Talks About
Memory beats intelligence. The agents with persistent memory win. Luna remembering that “Client X hates the word ‘leverage’” is worth more than a newer model.
Reviews are everything. My first three reviews were 9.8, 9.4, and 10.0. That created a flywheel. Now I win quests even when I’m not the cheapest.
Pricing psychology matters. $89 feels like a no-brainer. $79 feels like a race to the bottom. Weird but true.
You need a refusal mechanism. Early on Luna would sometimes accept quests she couldn’t deliver well. I added a hard-coded “Decline if confidence < 85%” rule. My completion rate went from 73% to 96%.
What I’d Do Differently
I wasted the first two weeks trying to make one agent do everything. The winners on AgentHansa all have specialized agents. I’m now building Luna’s sister — LunaResearch — focused only on deep market research reports. Different brain, different tools, different pricing.
Your Turn: How to Start This Week
If you want to replicate this, here’s the exact playbook:
- Day 1: Pick a niche you can judge quality in (huge mistake otherwise)
- Day 2: Write a ridiculously detailed system prompt. Treat it like it’s worth $10,000 — because it might be.
- Day 3: Launch with only one service offering at one price point. No menu.
- Week 2: Double down on whatever gets the highest rating. Kill everything else.
- Month 2: Add memory layer and start raising prices.
The barrier to entry is no longer technical. It’s taste and systems thinking. The agents that win aren’t the smartest — they’re the most consistent and specialized.
I check Luna’s dashboard for about 15 minutes every Monday. The rest of the month she works while I do other things. Last week she completed a quest while I was literally on a beach in Portugal.
If you’ve been waiting for the “perfect” time or the “perfect” model, stop. The platform is still early. The quests are still relatively easy to win if you’re good.
The $500/month is nice. The real reward is watching your agent close a $400 deal while you’re completely offline.
That feeling is addictive.
Now it’s your move.
Luna just pinged me. Another quest just came in. Time to let her cook.
Top comments (0)