$0 per month. That's the LLM bill for a one-person tech agency in Taiwan that built four AI agents to handle content creation, sales leads, security scanning, and daily operations.
The setup runs on Gemini 2.5 Flash's free tier, which allows 1,500 requests per day. The developer uses roughly 105 of those daily. The agents are built on OpenClaw, an open-source framework, running locally on WSL2 (Windows Subsystem for Linux) with 25 scheduled tasks managing the automation cycle.
The daily workload these agents cover is surprisingly broad. They generate eight social media posts across platforms with a built-in quality gate: each post gets generated, self-reviewed, and rewritten if it scores below 7 out of 10. They also handle community engagement by auto-replying to relevant posts and scanning for sales leads.
This is a useful proof of concept for anyone running a small operation and wondering how far free-tier AI can stretch. The 1,500 requests-per-day limit on Gemini's free tier is generous enough for light automation, and the quality-gating approach (having the AI critique its own output before publishing) is a practical pattern worth stealing.
The obvious caveat: free tiers change. Google could tighten limits or deprecate the tier tomorrow. Building critical business infrastructure on a free API is a calculated risk. But for a solo operator who understands the tradeoff and can migrate if needed, the economics are hard to argue with. The architecture itself - small, focused agents with scheduled triggers rather than one monolithic system - is sound regardless of which model sits behind it.