Run four OpenClaw agents for content, leads and ops on Gemini's free tier, with a quality gate and a hard cap on requests
A solo developer in Taiwan describes four OpenClaw agents on a home machine that draft social posts, reply to comments, research, scan for leads and monitor endpoints, using Gemini's free tier at about 105 of 1,500 daily requests.
Evidence: The author reports this. We have not checked it beyond reading the source.
The business problem
A one-person agency has no time to run content, lead follow-up and monitoring every day and wants to keep model costs near zero.
What was tried
Run four agents on OpenClaw on a home Windows machine with Linux subsystem, driven by 25 scheduled timers. Daily jobs generate eight social posts across platforms with a quality gate (generate, self-review, rewrite if the score is under 7 out of 10), reply to community comments for at most two rounds, research from RSS and the Hacker News API, run a security scanner as a lead magnet, monitor seven endpoints, and flag stale leads. To save tokens, agents never hold long conversations: each request reads pre-computed local notes, sends one focused prompt with all context injected, parses one reply and acts. Research collection uses plain HTTP with no model calls.
What was reported (mixed)
The author reports 27 automated Threads accounts with over 12,000 followers and 3.3 million views, about 7 percent use of the free request limit, and roughly $0 a month in model cost plus about $5 of infrastructure. No revenue or lead numbers are given.
Limitations
A self-reported Show HN post where a reader reported the linked repository returned a 404. Commenters questioned whether any revenue followed from the followers and views, whether 27 automated accounts is an acceptable practice on those platforms, and whether the output adds value. The author says an API key made from the wrong cloud project produced a $127 bill in a week, a loop bug burned 800 daily requests in one day, and a health check conflicted with the bot's polling and sent 18 duplicate messages. The author also says he uses AI to translate his writing.
What you need
An always-on computer, OpenClaw, a Gemini API key created in AI Studio, and comfort with scheduled scripts.
Sources
- Hacker News (Show HN by ppcvote) ↗ Firsthand write-up, published March 8, 2026
Source published: March 8, 2026. Last reviewed here: October 14, 2026. Spot a mistake? Tell us.
Tools in this workflow
- Gemini: The free-tier model behind the agentsTry Gemini
- OpenClaw: Runs the four scheduled agentsTry OpenClaw
Related workflows
Install an open-source library of 200+ go-to-market skills for Claude Code covering SEO, outreach, ads research and competitor tracking
A maintained repository packages go-to-market tasks as skill files for coding agents, including SEO research, lead finding and enrichment, outreach drafting, content, competitor tracking and social monitoring.
Turn a blog post into a Pinterest pin, alt text and LinkedIn post in about four minutes with a free Claude skill
A creator replaced an eight-step manual routine of separate prompts with a single Claude skill that produces the pin title, description, alt text and a LinkedIn post from a blog title and full text.
A 30-day test of posting one AI-generated image a day to LinkedIn that did not work: 1,709 impressions and 49 followers
A newsletter writer posted a ChatGPT-made image with a newsletter summary to LinkedIn every day for 30 days and reports poor reach, while one post offering free resources to a group did far better.
Film one long video and 20 short ones in a day, then repurpose them into a week of posts and a newsletter with Blotato and Claude Code
A creator-founder films one long-form video and 20 short videos on a single day, repurposes them into scripts, text posts and a newsletter, and distributes them with a scheduling tool and Claude Code.