The Project
A new distribution channel is forming, and most marketers have not noticed it yet. AI agents are starting to find, evaluate, and subscribe to software on their own, without a human in the loop, while the person who set them running is doing something else entirely.
MissingLinkz is the product I built for that market before most people accepted it existed: an agent-native marketing tool, built end to end with Claude Code, and marketed to its own customers with a markdown file. The customer is an AI agent.
The Shift: Agents Are Becoming Buyers
The signal came from watching OpenClaw, an open-source agent framework that went from zero to 250,000 GitHub stars faster than almost anything in recent memory. OpenClaw agents do not just answer questions. They take action, and they discover and install “skills” from a marketplace called ClawHub with over 13,000 of them. Agents browse it with semantic search, evaluate options, and, within a daily spending limit their owner sets, subscribe to paid services without asking first.
Then a concrete example made it undeniable. Postiz, a social scheduling tool competing with Buffer and Hootsuite, built a CLI installable via npm, added a SKILL.md so OpenClaw agents could discover it automatically, and returned structured JSON so agents could parse responses without a dashboard. Their founder was direct about the result: revenue climbed to over $45,000 in monthly recurring revenue. They did not build a better product. They built the same product with an interface agents could actually use, and agents chose it over competitors that have no CLI and no SKILL.md, because those tools are simply invisible to them.
That was the lightbulb: if agents are becoming buyers, what do they need that does not exist in agent-native format yet?
Finding the Gap
I mapped what AI agents actually do when they run marketing workflows. They write copy, schedule posts, send emails, generate images. But the moment they need a tracked link, the UTM parameters that let analytics attribute traffic to a campaign, they are stuck.
The numbers around this are uncomfortable: roughly 64% of companies have no consistent UTM naming convention, which makes about 22% of analytics data unreliable. So I looked at every UTM builder on the market: UTM.io, Bitly, Google’s Campaign URL Builder, CampaignTrackly. Every one has a web dashboard. Some have APIs. Not one ships a CLI installable from npm. Not one ships a SKILL.md. Not one returns structured JSON designed for agents. Not one has an MCP server.
The entire UTM builder category was invisible to AI agents. A universal marketing need, with zero agent-native solutions. That is where MissingLinkz came from. The name is a double play on the missing link in the agent marketing stack, with a Bigfoot mascot because Bigfoot is the original missing link.
What It Does
MissingLinkz started as a UTM link builder for agents and grew into something more useful. One command, mlz preflight, answers a single question: is this campaign link ready to publish?
In about two seconds it does three things. It builds a clean UTM-tracked link with enforced naming conventions, auto-lowercased, hyphenated, consistent across your taxonomy. It validates the destination, checking for 404s, SSL issues, redirect chains that strip tracking, and slow load times. And it inspects the landing page for social-sharing readiness: Open Graph tags, Twitter Cards, mobile viewport, canonical URL, page speed. Then it returns a structured go/no-go verdict, and if anything fails, it tells the agent exactly what is wrong so it can fix it or flag it to the human.
This is the preflight check that sits between “campaign ready” and “campaign live.” No pilot takes off without one. No campaign link should either. The Agent plan is $9 per month for 2,000 links, which works out at £0.004 per link validated, with a free tier of 50 links a month and no credit card, so anyone can test it first.
How I Built It
This is the part that matters if you are thinking about building your own products. I built the entire production system with Claude Code, not a prototype but the real thing: CLI, backend API, database, Stripe billing across four tiers, MCP server, marketing website, automated tests, CI/CD pipeline.
The process was sprint-based. I wrote a detailed project brief, then fed Claude Code one sprint at a time (core library, then CLI, then backend, then payments, then polish), each building on the last. Within a week of on-and-off development I had a production system: over a hundred passing tests, a live API, the CLI published on npm, and a SKILL.md security-verified on ClawHub. Total monthly infrastructure cost: under $30, because the backend runs on Cloudflare’s free tier.
The barrier to building agent-native products is very low right now. The hard part is not the code. It is identifying the right gap at the right time.
How Agents Find It
Traditional SaaS relies on Google, ads, and content marketing. Agent-native SaaS needs different discovery surfaces. MissingLinkz is listed on npm, where coding agents search for tools, and on ClawHub, where OpenClaw agents browse skills semantically. It ships with a SKILL.md bundled inside the npm package, so any agent that installs it immediately knows what it can do. The site includes an llms.txt file and an ai-plugin.json manifest, emerging standards that are, in effect, robots.txt for agents rather than search crawlers.
Here is the part I find most interesting. The SKILL.md does not just describe commands. It includes decision triggers that tell the agent when to use the tool, the business case with specific numbers, and a justification script the agent can relay to its human operator to explain why $9 a month is worth it. It is a product that markets itself to its own customers, and the marketing material is a markdown file.
Why It Matters for Marketers
Ten years ago, every business needed a mobile-friendly website or it was invisible to a growing slice of its audience. The same dynamic is forming with AI agents: every software tool will eventually need an agent-native interface. Google confirmed the direction in March 2026 by shipping a CLI for all of Google Workspace, built “agents first.”
For marketers, the implication is direct. The tools your agents use determine the quality of your marketing data. An agent that cannot create properly tracked links produces campaigns you cannot measure. And the market opportunity is wide open: this is like building a Shopify app in 2008. The positions are being claimed right now, and the cost of claiming one is close to zero. You need a clear idea, Claude Code, and the willingness to ship into a market most people have not noticed yet.
MissingLinkz is live at missinglinkz.io (npm install -g missinglinkz, free for 50 links a month). It is the first in what I expect to become a portfolio of agent-native marketing tools. When AI agents come looking for a tool in your category, will they find you?