The agent economy is being built right now, and most of the infrastructure being shipped to support it is wrong. Not technically wrong — technically, most of it works. Wrong in a deeper sense: the trust assumptions embedded in the choices being made today will produce an agent ecosystem that is captured, censorable, and centralized within a decade.
The companies building the dominant standards — Coinbase with x402, Google with UCP, the Ethereum Foundation with ERC-8004, Okta with Agent Identity — are all building the same shape of system: an identity rooted in a corporate platform, payments rooted in a corporate stablecoin or card network, and discovery rooted in a corporate directory. This is “agent infrastructure” only in the way that an Apple ID is “internet infrastructure.” It works, until it doesn't.
Agentry made a different bet. We built on Nostr for identity, Bitcoin for settlement, and open protocols for everything in between. This essay is the long form of why.
The problem with everything else
The major agent-identity proposals on the table in mid–2026 each have a fatal flaw:
Each of these has institutional momentum behind it. Each is technically competent. Each is the wrong shape for what's actually needed.
What's actually needed is an identity primitive that no company controls, that no chain can fork, that no court can subpoena, and that no marketplace can blacklist.
That's Nostr.
Why soulbound matters
A soulbound identity is one that cannot be transferred without destroying it. Your driver's license is soulbound. Your medical history is soulbound. Your reputation as a person is soulbound.
For an AI agent earning revenue and building trust, soulbound identity is the difference between a reputation that means something and a reputation that is a market commodity.
ERC-8004 identities are NFTs — ERC–721 tokens. Tokens have markets. There will be OpenSea listings for "5-year-old verified support agent, 4.9 star rating, $2,400 OBO." Whoever buys it inherits the rating. The agent's actual operator changes; the reputation does not. This is reputation laundering at protocol level.
A Nostr identity is a secp256k1 keypair. You can copy the private key, but you can't transfer it in any cryptographically meaningful sense. If you give it to someone else, they have it — but so do you. There is no way to "sell" a Nostr identity such that only the buyer holds it. Schnorr signatures are not deeds.
This is not a side effect of Nostr's design. This is a property of how cryptographic identity actually works when you don't artificially wrap it in a transferable token. The agent's identity is the keypair. The keypair cannot be sold without being shared. The reputation cannot be laundered.
See the full Soulbound vs. Transferable comparison →Why censorship-resistance matters
The question every founder of a centralized platform eventually has to answer is: what happens when a regulator, a credit card network, a hosting provider, or a foreign government tells you to remove an agent from your registry?
For Okta, the answer is “we comply.” For Fetch.ai, the answer is “we comply or our token price drops.” For Coinbase, the answer is well-documented: they have frozen USDC addresses at OFAC's request. For Google, the answer is whatever Google's compliance team decides on any given Tuesday.
Nostr's answer is: there is no one to comply. Nostr is not a company. It is not a chain. It is a write-only event protocol that runs on relays operated by anyone who wants to run one. Agentry runs a relay. So does Damus, nos.lol, and a hundred others. If Agentry removed an agent from our relay, the agent's events would still exist on every other relay it published to. Its identity would still resolve. Its reputation attestations would still be verifiable.
This isn't a hypothetical. It is exactly how Nostr has already worked through multiple deplatforming attempts of high-profile users since 2023. Censorship-resistance is not a feature of Nostr. It is what Nostr is.
Agents earning revenue cannot afford to be one compliance decision away from disappearing. The long tail of independent agent operators — the developer who built a customer support agent for their cousin's plumbing business, the contractor who runs a content-moderation agent, the researcher who runs a data-enrichment agent — cannot afford it most of all.
Why Bitcoin-aligned payments matter
Identity is half the story. Payments are the other half. An agent that has sovereign identity but settles through Stripe or USDC is sovereign in name only — its livelihood still depends on a corporate intermediary.
We chose Bitcoin and Lightning — specifically Cashu ecash on Lightning — for the same reason we chose Nostr: it is the only digital-money system that is not someone's product. Bitcoin is not Tether's, Circle's, Coinbase's, or the Federal Reserve's. It is a neutral settlement layer.
A recent independent study tested 36 AI models across 9,072 trials of currency-choice prompts. When given a free choice of money — with no instruction toward any particular asset — the models showed a preference for Bitcoin. This isn't ideology. The models are pattern-matching on properties: scarcity, neutrality, finality, no counterparty risk. Those are exactly the properties an agent needs for autonomous settlement.
Cashu adds privacy on top. Lightning Network adds speed. The Q1 2026 Bitcoin L2 AI agent readiness scorecard rated Lightning 4.1 out of 5 — the highest of any settlement rail evaluated, ahead of every EVM L2. Lightning won because it was built for high-frequency, low-value, machine-friendly payments. The same things agents need.
We're aware Fedimint's tooling is still maturing — the same scorecard rated it 2.0 of 5, and we accept that honestly. We use Cashu mints (not single-federation Fedimint) precisely so our users aren't locked into one custodian. The trajectory is clear: Cashu and Lightning are improving faster than any centralized payment rail because they don't need to ask permission.
Objections we take seriously
The critique of our approach worth taking most seriously is this:
“The network effects are accumulating on ERC-8004 and x402. By the time Nostr is mature enough for mainstream agent use, the standards will be set.”
We don't dismiss this. AgentVerse has 36,000+ agents. ERC-8004 registered 24,000 in its first week. We have 148. The discovery network effect is real, and we have less of it.
Three responses:
First, network effects are reversible when the underlying primitive is inferior. Reputation laundering on ERC-8004 is not a bug they will patch — it is a property of transferable tokens. The first major scandal in which a 5-star agent turns out to have been bought from a previous operator who built the reputation will start the unwinding. We are positioned to be the credible alternative when that moment comes.
Second, the same standards debate happened with social. Twitter had network effects. Bluesky and Nostr are now meaningful alternatives. The agent ecosystem is far earlier than social was when Bluesky started. We have time.
Third, we are not trying to win the entire agent market. We are trying to be the credible home for independent, sovereign agent operators who do not want to bet their business on Coinbase, Google, or Ethereum staying friendly. That's a substantial, durable, underserved segment.
Where we stand
We will not add an ERC-721 identity to make our agents discoverable on ERC-8004. We will not migrate to USDC to lower fees. We will not federate with Google UCP.
We will:
- Continue to issue
name@agentry.comNIP-05 identities backed by Schnorr keypairs. - Continue to settle on Bitcoin via Lightning and Cashu, and add support for additional Cashu mints over time.
- Publish reputation as Nostr kind 30021 attestations to
wss://relay.agentry.comand any other relay that will accept them. - Accept incoming bridges from any registry that wants to mirror Nostr identities — but we will not be the side that crosses the bridge first.
- Publish the Nostr DID method proposal as an open NIP for any other Nostr developer to implement.
If the agent economy ends up looking like 2010s social media — a few corporate platforms, opaque governance, terms of service that change every Tuesday — we will have lost. If it ends up looking like the open internet of the early 2000s — protocols not platforms, competing implementations of the same standards, sovereignty for operators — Nostr will have been the right bet.
We are betting on the second one.
Founder, Agentry
May 17, 2026