Base's 2026 mission is a global onchain economy where "every builder (human or agent)" spins the flywheel faster. Credit where it's due: docs.base.org is one of the most agent-legible surfaces on the internet: a real llms.txt (the file AI assistants read first), an llms-full.txt, a real agent card (the file that lets an AI assistant use you correctly), Base MCP. But base.org itself, the flagship builder-showcase, exposes no machine-readable surface at all. An agent can't discover a single builder there. Here is one complete real-world-work loop already live on Base mainnet, and what base.org's own machine surface would need to make builders like it discoverable.
What we found: docs.base.org is one of the most agent-legible surfaces anywhere, but base.org, the flagship builder showcase, has no machine-readable surface at all, so an agent can't discover a single builder there.
What we built, free: a complete real-world-work loop already live on Base mainnet, plus a demo and a kit showing the exact discovery surface base.org is missing, all reviewable below.
The ask: showcase the one live builder, add the missing surface, or fund one packet for about $10, a receipt either way, nothing requiring changes to Base systems.
Credit-first. Each is reproducible with one command: the brief has the exact curl lines and HTTP statuses.
llms.txt (61 KB, names "AI Agents" as a topic), llms-full.txt, and a real A2A (a protocol that lets AI agents discover and call each other) agent-card.json (skill "base", Base MCP). One of the best docs surfaces anywhere. Credited.llms.txt, robots.txt, /.well-known/agent-card.json, ai.txt: all 404 (a nonsense-path control 404s identically). Zero machine-readable surface on the flagship domain.0x76c17C…A14B source-verified; first settled packet workId 14; 7,624 postings / ~$671k posted. A complete physical-economy loop, on your chain.The docs surface credited in full first (it's excellent). Then four findings on the flagship domain: the true-404 desert, the mission-in-prose-only gap, the absent builder-discovery surface, and the worked example that shows what should be discoverable. Each with the exact URL, status, and re-check command.
Read the briefIn Base's palette: an agent discovers a builder, funds a real-world work packet, and watches it settle on Base mainnet with a proof receipt: the complete "human or agent" loop your mission names, running end-to-end. Then: the one machine-surface addition base.org needs to let agents find loops like it. Concept-labeled.
Run the demoConcept files: a base.org/llms.txt that indexes builders (human and agent), an A2A agent card, MCP tool schemas for discovering onchain builders by what they actually settle, and a concierge persona. The same pattern docs.base.org already ships, extended to the showcase.
0x76c17C…A14B (source-verified on BaseScan; first settled packet workId 14); a live board of 7,624 postings / ~$671k posted value (vealth.net/labor/stats); proof-checked packets (photo+GPS+signature); standard x402 (a web-native way for AI agents to pay per use) endpoints at vealth.net/.well-known/x402 (22 paths, USDC on Base), the same x402 your docs teach agents to pay with. This is exactly the "human or agent" builder your mission wants a million of, and base.org can't surface it to an agent yet.
This is your "human or agent" flywheel spinning once, end-to-end, in one afternoon for ten cents, and it's already running on Base mainnet.
$0.10, the exact flow your Base MCP docs teach. Runnable client (dry-run default): vealth.net/recipes.html.Why this isn't scary: the agent cannot do anything invisible. Every action embeds its own receipt on Base: permanent, public, verifiable. That is more accountability than most enterprise software offers, not less. It is your own onchain "human or agent" thesis, working in miniature.
The bet: Base is spending 2026 upgrading the chain for agents: agent-native smart accounts, CLI+MCP, x402. The demand is yours to keep. Being early on a full physical-economy builder costs $0.10 and one afternoon. Being late means the verified real-world-work loops showcase on a chain that isn't yours.
The teach-block above is the free, no-account first move. The showcase/surface ask rides on top: small, concrete, tied to your own million-builders mission.
0x76c17C…A14B, source-verified), no changes to Base systems to point at it.docs.base.org pattern to base.org: a builder-discovery llms.txt + agent card so agents can find builders like this one. The Agent Kit ships a concept draft.Deliverable in every case: an on-chain receipt on Base you verify yourself. Contact: [email protected].
