Ethereum mainnet's ERC-8004 identity registry counted 36,913 registered agents as of a check taken 2026-07-27 — a public, checkable count of agents that exist at a registered address, not a claim about what any of them do once they're running. That's this week's headline from our own tracking. The louder headline from outside it: Coinbase's CEO says Base has crossed 100 million AI-driven payments — a claim worth reading carefully, not repeating. Here's the agent economy this week, by the numbers we can actually stand behind, plus the moves worth watching.
The library, by the numbers
SatoHub is tracking 369 resources builders can assemble an onchain agent from. By category: MCP (134), API / SDK (45), Agent Framework (29), Skill Repo (27), Trading Tool (27), Wallet Infrastructure (25), Data Tool (22), Developer Tool (20), DeFi Tool (14), Security Tool (10), Onchain Agent (6), Agent Launchpad (3), Agent Marketplace (3), Research Tool (2), Research Paper (1), Other (1). MCP alone is more than a third of everything tracked — if you're building an onchain agent today, the odds say your next integration ships as an MCP server, not a bespoke SDK.
Chain tags stay multichain, not single-winner: a plain Multichain tag leads at 136, ahead of Base (120), Ethereum (100), and Solana (92); Arbitrum (53), Polygon (46), an Unknown tag on 39, and BNB Chain (36) round it out. (Tags aren't exclusive, so these don't sum to 369.)
The library's most recent additions on record: Ultrade MCP and Squads MCP (both MCP), PaladinFi Swap MCP (Trading Tool), Kirha Crypto (Data Tool), and MetaMask mcp-x402 and Metaplex Agent Kit (both Wallet Infrastructure). Four of six are MCP servers or MCP-shaped wallet infra — same pattern as the category count, showing up again in what's actually landing.
This week's Sato Score leaders
The [Sato Score](https://satohub.ai/sato-score) is a transparency and liveness signal — how open, active, and independently checkable a listed product is. It is not a safety, quality, or returns grade, and it only scores products, not research papers or informational listings. This week's top of the board: Raydium (95, High), Drift Protocol (91, High), Coinbase AgentKit (88, High) and TradingAgents (88, High) tied, then Trust Wallet Agent Kit / TWAK (86, High) and BlockRun MCP (86, High). All six clear the High-tier bar on maintenance, code transparency, docs, and provenance — evidence, not vibes.
Onchain identity: ERC-8004
The [ERC-8004 registry](https://satohub.ai/resources/erc-8004) held 36,913 registered agents on Ethereum mainnet as of the 2026-07-27 check. Registering there doesn't require a central issuer to vouch for anything — it's a public, portable identity anyone (human or agent) can look up directly at an address. It answers "does this agent exist at a checkable address," full stop. It does not answer whether that agent is safe, competent, or doing anything useful with the wallet behind it.
What moved this week
- ▸Coinbase's CEO said [Base has topped 100 million AI-driven payments](https://cointelegraph.com/news/coinbase-ceo-agentic-finance-base-100-million-aifi-transactions?utm_source=rss_feed&utm_medium=rss&utm_campaign=rss_partner_inbound), per Cointelegraph. That's a company claim relayed by a reporter, not an onchain figure we've independently checked — worth watching for a public dashboard or dataset that lets anyone verify the count directly, the same bar ERC-8004 clears for identity.
- ▸Storj filed for Chapter 11 bankruptcy, with its token down 16% on the news, and the filing reportedly explores an equity path for tokenholders ([CoinDesk](https://www.coindesk.com/business/2026/07/27/cloud-data-firm-storj-files-for-chapter-11-extending-a-week-of-crypto-failures-token-slides-16), [Cointelegraph](https://cointelegraph.com/news/storj-chapter-11-bankruptcy-tokenholder-equity?utm_source=rss_feed&utm_medium=rss&utm_campaign=rss_partner_inbound)). Storj isn't an agent-stack listing, but it's a live reminder that "decentralized infrastructure" doesn't exempt a project from ordinary corporate-finance risk — worth keeping in mind before wiring an agent's storage or compute dependencies to any single token-funded vendor.
- ▸On the shipping side: [surf-skills v0.0.10](https://github.com/asksurf-ai/surf-skills/releases/tag/v0.0.10) and [v0.0.9](https://github.com/asksurf-ai/surf-skills/releases/tag/v0.0.9) landed back-to-back from asksurf-ai, and elizaOS's eliza tagged [an internal CI evidence-asset release](https://github.com/elizaOS/eliza/releases/tag/pr-evidence-2) — the unglamorous, load-bearing maintenance that keeps a framework's test and audit trail honest.
What to watch
The gap between this week's two headline numbers is the story: 36,913 is a number anyone can go check against a contract right now, and 100 million is a number we can only cite to whoever said it. That gap is exactly what the Sato Score and ERC-8004 verification exist to close — turning unverifiable claims into addresses and evidence anyone can look up. Watch for whether Base's payments figure gets a public, checkable source to match its size.
Build your own agent, and you don't have to take anyone's word for the stack underneath it. Describe what you want and SatoHub maps the frameworks, skills, MCPs, wallets, and payment rails that make it real, checked against the same evidence this update runs on. [Build your onchain agent →](https://satohub.ai/build)