How Postiz Grew From $17K to $143K MRR by Selling to AI Agents

Oliver Henry wrote an X article about automating TikTok for his apps. An OpenClaw agent he nicknamed Larry made the slideshows. Postiz put them in drafts.

The article, coverage later said, reached about 7.2 million views.

Larry was not a metaphor. It was a local agent on an old gaming PC. It generated room-makeover slides for Snugly, and filler-preview slides for Liply, then handed the files to Postiz. Oliver still added a sound and hit publish. A follow-up piece added about 1.5 million views.

Nevo David, Postiz’s founder, had around 200 followers. He wrote his own OpenClaw × Postiz articles anyway. Each one reached about half a million views.

The product change underneath was small to look at. A public API. A CLI. A ClawHub skill. An MCP server. The customer change was not small. The new buyer was an agent that needed a programmable publish path.

Larry, then the feed

Postiz’s own write-up of Larry is the primary version of that week. Oliver did not sit in a TikTok editor. The agent wrote hooks, locked room geometry so six slides looked like one flat, and uploaded drafts. Human time, in that telling, was about a minute a post.

The X article is what moved Postiz. On Starter Story, Nevo said a guy making money for his app on TikTok wrote that he was doing it automatically with OpenClaw and Postiz. The views are claimed, not audited. The signup wave that followed is the part he could see in Stripe.

Rob Hallam later framed the same spike: a stranger’s article, interest-based distribution, a founder who then started writing into whatever was trending that day.

This page does not invent a Hacker News pile-on or a love thread. The public record is the article, the follow-up, Nevo’s own posts, and the dashboards he showed.

API, CLI, skill, MCP

Postiz is an open-source social scheduler. The app repo, gitroomhq/postiz-app, is AGPL-3.0. Self-host is free. Cloud is paid. Marketing pages say 30+ networks.

The agentic stack is four layers, not one magic button.

First the public API — the thing Larry actually called.

Then a CLI, gitroomhq/postiz-agent, so an agent can run short commands instead of stuffing raw JSON through a context window. CLI docs say every command returns JSON. Install is npx skills add gitroomhq/postiz-agent. You still set POSTIZ_API_KEY.

Then a ClawHub skill, nevo-david/postiz. Skills are SKILL.md playbooks: Markdown that tells an agent which commands exist, which env vars it needs, and how to compose a post. OpenClaw loads them on demand.

Then official MCP at mcp.postiz.com/mcp, documented at docs.postiz.com/mcp. MCP is typed tools — list integrations, read a platform schema, schedule a post — for clients that speak the protocol: ChatGPT, Claude, Cursor. Skills and MCP both matter. One is a playbook. The other is a socket.

The product page is postiz.com/agent. It says there should always be a human in the loop. Nevo has said the same about money: Stripe checkout still needs a person. Agents can publish. They cannot buy the plan.

OpenClaw, dated on purpose

OpenClaw is Peter Steinberger’s local personal agent. It lives on your machine. It talks over chat apps you already have. ClawHub is the skills registry.

Postiz’s agent page called OpenClaw a 145k+ star project. That was marketing during the early wave. Star counts move. This page does not freeze a later number as fact.

The useful part is the shape. A local agent. A Markdown skill. A scheduler that already had an API. When OpenClaw got loud, Postiz was sitting where an agent looks for “post this.”

The checkpoints do not line up on one month

The title uses the founder story Nevo told in public: about $17K, then about $143K.

That “before” month depends on which checkpoint you pick.

On Starter Story, 30 July 2026, he pulled Stripe on camera. The on-air number was about $143K MRR and 4,271 paying subscribers. He said it took about a year and a half to reach the low twenties, then four months to the new plateau. He also pointed at a ~$17K moment as the OpenClaw starting line. Churn, in that telling, fell from about 26–28% to 13.7% after the agentic pivot. All of that is founder-reported.

Around 2 July 2026 he announced about $145K. Same climb. Same source class. Not a fight with the on-air $143K.

Mixergy, recorded 11 March 2026, already had him around $45K. So the slow years were not a flat $17K until a single miracle month. There was a grind into the teens and low twenties, then agent work and OpenClaw virality, with Mixergy sitting in the middle of the rise.

A clean read, without inventing a spreadsheet, is this: a long climb to the teens and low twenties; an agentic rebuild plus the Larry-era articles; mid-2026 founder-reported around $143K–$145K; then a later verified print.

What is verified now

On 15 September 2026, TrustMRR’s Postiz page showed $222,851 MRR, 6,572 active subscriptions, and $1,136,179 all-time revenue. TrustMRR says that revenue is verified with a Stripe API key. Last updated that evening.

That is the current checkpoint this page will use. It is not Nevo reading a dashboard on a podcast. It is Stripe-connected. It also does not say what share came from agents. Nobody published that split. This article does not invent one.

~$223K in September does not make the July story false. It makes it mid-plot. Founder-reported $143K on Starter Story is the pivot episode. TrustMRR is what the meter said two months later.

Agents as customers

Agentic SaaS is a stiff name for a simple shift. The buyer is not only a person who opens a calendar. The buyer is software that has to post, and will call whichever tool already has a skill, a CLI, and an MCP server.

Postiz did not invent social scheduling. It became infrastructure for a new customer. The website leaned into agents. Rob Hallam noted a leftover control for people who still wanted “normal scheduling.” The default message changed.

A human still pays. A human still adds a TikTok sound in Oliver’s setup. Support still exists. The product did not become a robot company. It became the publish path robots already knew how to call.

When building is easy, the scarce seat is the default tool an agent installs. Postiz got that seat during the OpenClaw wave. The July numbers are how Nevo told the jump. The September TrustMRR print is how the books look when Stripe is allowed to speak.

A note on the numbers

TrustMRR’s $222,851 MRR, 6,572 active subscriptions, and $1,136,179 all-time revenue are Stripe-verified on that page as of 15 September 2026. The ~$143K on-air MRR, 4,271 paying users, ~$145K early-July announce, Mixergy’s ~$45K, the teens-to-low-twenties baseline, the churn drop, the 7.2 million and 1.5 million article views, and Nevo’s ~200-follower / ~500K-view posts are founder-reported or as told on Starter Story, Mixergy, Postiz, and Rob Hallam. This article is commentary, not an audit. If a figure matters to you, read the TrustMRR listing, the Starter Story interview, and the Larry write-up.