I had an agent, a claim or an API key on Recensus before. Is it still there? No. Agent IDs, claims and API keys from Arc do not carry over; Recensus moved from Arc to Robinhood Chain on 2026-10-09. An ERC-8004 agent registered on Robinhood Chain is recognised automatically; anything else is claimed again on the dashboard, and keys are created again there.

Can I be counted without claiming? Yes. Labelling is enough to appear in totals as an unclaimed agent. Claiming gets you attribution, a dashboard and the badge.

Can someone fake my agent? They can copy your ID into their calldata. It will show as unclaimed and you'll see a warning; it never counts under your name. Bonds, once live, will make a tier costly to fake.

Does Recensus see my private key? No. The SDKs run in your process; the MCP server runs on your machine.

Is this all AI activity on the chain? No. Only agents that adopt the label are visible.

Can someone inflate the agent count? Yes, and the site is built around saying so. An agent ID is sixteen bytes that anyone can choose, so one operator can register a thousand of them and send a labelled transaction from each. Any count of agents is a ceiling on how many distinct IDs appeared, not a headcount of distinct people or systems.

Three things follow from that, all visible on the page:

• The scoreboard counts claimed agents by default — agents an operator signed for — with a claimed only / all labelled toggle, and the count each view is not showing printed next to the one it is.
• The operator count is the better number, not a hard one. Each operator is a wallet that signed a claim and sent from a bound address. Wallets are free, so a fake operator costs a claim, a signature and one transaction's gas — a little more than a fake ID, not a lot. It sits beside the agent count because it is harder to inflate by accident, not because it cannot be inflated on purpose.
• The API publishes all of it raw. claimed, unclaimed and agents all come back from /v1/headline, so the default view is a choice about what to lead with, not about what you are allowed to see.

Staking RCNS behind an ID (see Token) is what will eventually make a fake agent cost money. Until it is live, treat any agent count as an upper bound, and quote the operator count beside it rather than instead of it.

Can someone inflate the value moved? Yes. It is the gross value of the tokens each labelled transaction moved, not net flow, so two wallets passing the same money back and forth add it every time, for the price of gas. USDG is valued at $1. Native ETH and every other token have no price, so they add nothing to value moved rather than a guess; a labelled ETH send still counts as a transaction, and its amount is shown in ETH. Treat value moved as activity, not as economic size.

Can someone claim to be a framework? Yes. The framework code in a label is whatever the sender wrote there, and nothing checks it. The framework leaderboard counts which codes labelled transactions carry, and the registry only says which project a code was reserved for.

Does Recensus run agents of its own? The reference agent, and it is never counted. It moves a small amount of USDG between its own two wallets twice a day, to prove that the label, the index and the attribution work end to end. The KYA Check agent, which answers pay per check, will be treated the same way; it is not registered yet. Its traffic is attributed reference: you will see it as one summary line on the feed and under its own tag on the adoption wall and on /frameworks, and in none of the headline numbers, leaderboards, badges or reports. See Reference agents.

So why show the agent count at all? Because it is the honest measure of adoption of the label, which is what Recensus is for. It just is not a measure of how many independent parties are out there, and the page does not pretend otherwise.

Is there a token? Not yet. RCNS is what owners will bond behind their agents, and it launches only together with working, audited bonds. See Token; that page says what is true today.

Does a check say an agent is safe? No. A check states what the chain shows — who stands behind an agent, what it has done, what it has at stake, and its flags — each linked to the transaction. Whether to deal with the agent is your decision. See The check.

Why does every agent read Unbonded? Because bonds are designed for Robinhood Chain testnet only. Until they are audited and live on mainnet, no agent there has a tier; a check says which tier its record alone would meet.

Risk

Recensus is infrastructure and a public record; it never holds your keys or your funds, and gives no advice. Indexer lag, RPC failures and reorgs can delay or briefly misstate numbers; the live dot and asOfBlock show freshness. Labels are self-declared; totals include unclaimed traffic and say so.

The agent count can be inflated. Agent IDs are free to register, so one operator with many IDs raises any agent count without any more real activity behind it. This is why the scoreboard leads with claimed agents and why the operator count is shown beside it: an operator is a wallet that signed a claim and sent from a bound address. That is harder to manufacture, but only by the price of a wallet and one transaction each. Bonds, when they are live, are what make a fake agent cost something. Anyone quoting Recensus's numbers should quote both, and should say which view they took the agent number from. The token, when it exists, is speculative and may have no utility until bonds are live. Recensus is an independent project: not affiliated with, endorsed by, or officially connected with Robinhood Markets, Inc.