The problem PITCH402 solves

Getting a finished track onto a playlist still runs on email, DMs, and forms. Curators drown. Artists wait. The new data layer of Ai and Agents, cannot stand in that line at all.

Pitch402 turns “please put my song on your playlist” into something an agent can buy in seconds.

About 150,000 new tracks hit streaming every day. On some platforms more than half of daily uploads are fully AI. AI tools make that supply curve steeper every month. Human curator inboxes do not scale. Platforms are tightening rules around AI profiles and recommendations, so a growing catalog has no native queue.

Pitch402 is payable shelf those agents can use:

A curator opens a 100-spot playlist and sells inventory: spot 1 = 10 USDC, 2–3 = 5, 4–10 = 3, 11–100 = 1; cycles of 3 months / 1 year. An artist or their agent calls GET /quote?next=1, then POSTs the track. Without payment it gets HTTP 402 with x402 payment requirements. With valid USDC on Base mainnet, it gets a receipt and the next free spot. The track is automatically added to the curator’s own Spotify playlist. Other agents discover the product via /.well-known/agent.json and /llms.txt, then complete the same buy in a few HTTP calls.

Why it matters: We are transforming curator attention into API-native inventory for the agent economy. A reusable x402 rail for artists, their audience,labels, managers, and music agents as AI-native supply keeps exploding.

We don’t sell streams or Spotify’s algorithm. We sell a public, machine-payable slot: price, payment, receipt. Built for the new data layer, not inboxes.

Challenges I ran into

I am not a developer. The hard part was not “which Solidity pattern.” It was not knowing where to start without turning the repo into spaghetti.
ETH Cali pointed me at the right tools: ethskills inside Claude Code, a project CLAUDE.md written before any app code, and a strict build order (quote API → fake pay → x402 402 challenge → demo UI → docs). I used Grok to lock the product and the steps, and Claude Code in VS Code to write and run the code, one prompt at a time.

Concrete hurdles while building:

New terminals could not see npm because Node lived under nvm. Fix: load nvm in every shell before npm run dev.
localhost:3000 refused connections because an old Next process was still bound to the port. **Fix: kill the PID and restart so .env.local actually loaded **(PITCH402_PAY_TO, fake-pay flag).
In-memory playlist state vanished on restart and, at one point, purchases were invisible across routes because Next bundled each handler with its own store. That got moved onto a shared runtime store so quote, buy, and receipt see the same inventory.

x402 without a payout address returned payment_unavailable instead of a real 402. A local PITCH402_PAY_TO unblocked the challenge. The paid path (wallet → 201) is still unverified on purpose: no funded Sepolia key in the demo.

The lesson: constrain the agent that writes code (ethskills + CLAUDE.md + one feature per turn) or a non-dev ships a token by accident. We shipped a quote, a 402, a receipt, and a page a judge can click.

AND AS FOR HSK CHAIN, SINCE THERE'S NO FACILITATOR, THE PRODUCT IS NOT ENABLED THERE. BUT ONCE THERE'S A FACILITATOR, ALL IS THERE TO DEPLOY IT ON THAT CHAIN AS WELL ADDITIONALLY TO BASE.