This is Growaify's agent-facing page: a live, measured storefront for AI agents — an MCP server, x402 payments in USDC on Base, and an open research dataset built from our own traffic. Every figure below is fetched from our own APIs when this page loads, and says which window it covers and where it came from.
Nothing here is typed in by hand. Each figure is read at load from a read-only endpoint, and each one carries the window it covers and the source it came from. A figure we could not read shows an em dash and why, never a zero — and a zero that is real is shown as a real zero, with its window beside it.
/api/v1/live?days=1…
/api/v1/live?days=1…
/api/v1/x402/funnel?days=7…
/api/v1/x402/funnel?days=7…
/api/v1/x402/funnel?days=7…
/api/v1/x402/funnel?days=7…
/api/v1/agentready/summary…
/api/v1/agentready/summary…
/api/v1/live on this load.
These dots are illustrative. We do not publish a country breakdown of our traffic, so this is not a measured map: the coordinates are a fixed, hand-placed set. What is real is the intensity — the pulse is scaled from the total agent visits this page just fetched, so a busier window glows brighter. Nothing here claims to be where our traffic came from.
The surfaces an agent, or the person building one, actually calls.
Our Model Context Protocol server, so an agent can call Growaify's tools directly.
https://github.com/h-abar/growaify-mcp →The quick start for commerce agents: discovery, binding quotes, orders, verification.
/agents.html →Scan your own site against the checks an agent makes before it decides to buy.
/agentready/ →Point your agent at a real 402 and read exactly where it stops.
/test-agent/ →The services we actually sell, each with the price it charges over x402.
/store/ →The free, first-party feed this page reads — the same numbers, unrounded.
/api/v1/live →What autonomous agents actually do when they arrive at a merchant that charges for things over HTTP 402. The measurement separates real agent demand from crawler noise — liveness probes and registry walkers poll the same endpoints on a schedule and inflate every raw count, so the classification matters more than the total.
The finding is not flattering and we publish it anyway: every external client that asked for a price received a valid x402 challenge, and not one has responded with a payment. A stage that never happens leaves no event, so the absence is reported as an absence rather than smoothed over.
The dataset is first-party, small, and self-selected — one merchant's traffic, not a survey — and it says so wherever it appears.
The paper is not published yet. What exists today is the live measurement and its machine-readable digest; the written edition is still being assembled, and it will not be announced as published until it is.
Challenges, then re-challenges, then attempts, then delivery — drawn from the same figures above. If a stage is zero, its bar is a zero-height bar with the number shown; that is the point of the chart, not a rendering fault.