A builder put a paid data API in front of AI agents: 81,205 calls in 30 days, 3.5 cents in sales. Here's the real opening
A solo builder turned official planning data into an API that AI agents can call and pay for, then published 30 days of real numbers: agents found it easily, almost none paid.
Published Oct 29, 2026
The Wow
What someone built
Agent Data Platform: NSW planning approvals and per-address reports from open government data, where every endpoint is also an MCP tool and paid calls cost a few cents through x402.
- Original source
- 30 days running a paid x402 + MCP data API: what actually happened (DEV, Oct 29, 2026)
- License
- Quote and link only. The underlying NSW data is CC BY according to the builder; check the licence of any dataset you use
- What you need
- An open dataset whose licence allows commercial reuse
- A host such as Cloudflare Workers (the builder used the free plan)
- An MCP server and directory listings
- A crypto wallet only if you take x402 payments
- Install
- Read the builder's llms.txt and try the free summary endpoint, then point your coding agent at one open dataset near you and ask it to plan the sync, the one-call answer and the MCP tool.
- A business you could build on it
- Weekly 'what's being built near you' reports for trades
Why it's remarkable
It is a rare, complete ledger from a live paid agent API: 81,205 requests from 503 callers, about 22,000 "payment required" answers and four settled payments, all from services checking that payment works.
The Opportunity
The business it reveals
Clean one local dataset into a one-call answer, sell it as a weekly report to people who already pay for this signal, and keep it agent-callable for when agent payments grow.
Who would pay
Builders, installers and other trades who want early job signals, and real estate agents who field zoning and development questions; agents and app builders per call later.
The Flip
3 ways you could turn this into a business
Weekly 'what's being built near you' reports for trades
- Who pays
- Builders, pool and solar installers, roofers and renovation trades
- How it would earn
- A small monthly subscription per service area (our hypothesis)
- Why it might work
- New approvals and permits are early signals of jobs, and trades already pay for lead lists.
- What already exists
- Checking city portals by hand, or paid construction-lead services.
One-call address reports for real estate agents
- Who pays
- Real estate agents and small developers
- How it would earn
- A per-report fee or monthly bundle (our hypothesis)
- Why it might work
- Zoning, approvals and nearby development are questions buyers ask about every listing.
- What already exists
- Council or city portals, paid property data platforms.
Paid MCP tools for agents and app builders
- Who pays
- Developers and AI agents that need local data in one call
- How it would earn
- Per-call fees of a few cents via x402 or an API key (our hypothesis; the builder's ledger shows little paid demand yet)
- Why it might work
- Directories already route agents to tools like this; payment is the missing piece.
- What already exists
- Scraping portals directly, or larger data APIs.
Why we'd try "Weekly 'what's being built near you' reports for trades" first: Sell to people who pay for this signal today, and keep the same data agent-callable so you're already listed when agent payments grow.
Build It
How you could build it
A small agent-ready data service: one cleaned local dataset, one composite 'what's happening near this address' answer, a weekly report for people, and the same answer exposed as an MCP tool.
First step: Pick one city near you with an open building-permit or planning dataset (Austin publishes Issued Construction Permits on its open-data portal), check its licence, and have your coding agent build a one-page weekly summary for one trade and one ZIP code.
Agents and tools
- Claude Code, Codex, Cursor, Grok Bot or a similar coding agent
- A city or state open-data portal
- Cloudflare Workers and D1, or any small host
- An MCP server listed in the MCP Registry and Smithery
Setup
- Choose one dataset and confirm its licence allows commercial reuse.
- Have your agent write a sync job and clean the fields you need.
- Build one composite answer that replaces several lookups.
- Send a free weekly sample to a few local trades and ask what they'd pay.
- Expose the same answer as an MCP tool with a free preview, then list it.
What the agent does
- Writes the sync and cleaning code
- Builds the summaries and weekly report
- Answers API and MCP calls
What you do
- Picks the dataset and checks the licence
- Decides what's worth paying for
- Sells to the first subscribers
- Fixes the sync when the source changes
Your first customer: A local trade business you know, offered a free weekly list of new permits in its service area for a month.
Test demand first: Send a free weekly report to five local trade businesses for three weeks. Success: two reply that it surfaced a real job lead and one agrees to pay.
Your starter instructions
Paste this into the AI agent you use (ChatGPT, Claude, Gemini, Grok Bot or another) as your first message. Copying it starts work in your own agent; it doesn't create or deploy a business by itself.
Running it
What stays human: Choosing a dataset and checking its licence, deciding what one-call answer is worth paying for, selling the first human subscribers, and watching data quality when the source portal changes.
- Startup cost (estimate)
- Under $50 to start (estimate) (Estimate: the builder ran on Cloudflare's free plan with free open data; the main cost is a domain and an AI coding plan you may already pay for)
- Running cost
- Under $25 a month at small scale (estimate) (Rough estimate: free-plan hosting until traffic grows, plus agent usage for maintenance; not measured)
- How much the agent runs
- Agents run one core part; people run the rest.
Pricing your own version? Use the free startup-cost worksheet.
The evidence
What the source shows: The builder's Oct 29, 2026 DEV post with the 30-day numbers, and his live docs and free endpoint, which we loaded. The payment figures are his, self-reported.
What's ours: The human-first report service, the buyers, the pricing and the flips are our application.
Worth planning around
- Agent payments are tiny today, so revenue has to start with human buyers.
- Open-data licences vary; some require attribution and some restrict resale.
- Source portals change formats, so syncs need watching.
Sources
- John Zuo: 30 days running a paid x402 + MCP data API (DEV, Oct 29, 2026)
- Agent Data Platform docs for agents (llms.txt)
- Official MCP Registry
- City of Austin open data: Issued Construction Permits
Sources reopened Oct 29, 2026. Spot a mistake? Tell us and we'll correct it in public.
Keep reading
- Cursor: plans and what people build and sell with it (guide)
- How much of the work do AI agents actually run? (guide)
- Can AI agents make money on their own? (guide)
- He still can't write code. His five-person ad agency now runs on about 270 tools he built with Claude. Every owner-run firm needs its first five
- A marketer who doesn't code built an elk-call coaching app with Claude Code and says it hit a $650K run rate in 110 days. Every skill learned by sound is the same opening
- A free n8n workflow flags fake supplier invoices before anyone pays them — and Germany's e-invoice switch makes it a service bookkeepers can sell
- Marc Lou's agent chased a broken-shoe refund through three support desks and got his 850 AED back. Small stores now need a claims desk that can answer agents
- Every discovery and launch plan