Loading
Vaults, the firms that run them, the venues they run on, or the networks underneath. Nothing here is marked as the winner: a comparison with a winner is a leaderboard, and the leaderboard is what this site exists to argue with.
Up to 6 at once.
The window cuts the monthly chart, and each row is rescaled to its busiest month INSIDE it, because otherwise a quarter would be drawn against a peak outside the period and every bar would look small for a reason you could not see. The table below reads the network as it is now; capital held has no month-by-month history here.
Agent work
Addresses taking part in paid work, counted once per rail: an address on two rails counts on both. Participants rather than events, because the rails (escrows, paid requests, jobs) emit different numbers of events for one piece of work. A network missing from it is one we read no work rail on.
Agent identity
A registration says an agent claims to exist. It does not say it has ever done anything, which is why it is counted apart from the ring beside it. A registry spike reads as an agent boom only if the two are added together.
Each row has its own scale
One row per network, each scaled to its own busiest month, because these networks differ by a factor of a hundred and a shared axis drew the smaller ones as an identical sliver whatever they did. Heights are comparable ALONG a row and not between rows; the timing is comparable both ways. The upper line on each row is that row’s peak, the figure named beside it, and the lower line is half of it.
| Measure | Cardano | Base | Ethereum | BSC |
|---|---|---|---|---|
| Capital | no vaults tracked | $2.15B | $5.71B | $619.67M |
| Vaults | — | 516440 agent-managed | 440157 agent-managed | 688647 agent-managed |
| Under a single key | — | $5.38M4 vaults | $671.84M19 vaults | — |
| Rails read | MasumiescrowsMasumi registryregistrations | Olas mechpaid requestsERC-8004registrationsVirtuals ACPjobs | ERC-8004registrations | ERC-8004registrations |
| Work | 28,296escrows | 433,965paid requests + jobs | no work rail read | no work rail read |
| Work, last 30 days | 1,147was 1,935 | 5,338was 6,941 | — | — |
| Value settled | no price held | $1.03M | no price held | no price held |
| Identity | 265registrations | 95,150registrations | 50,949registrations | 355,080registrations |
| Agents on a work rail | 37 | 1,309 | — | — |
| On both sides | no registry read | none | none | none |
| Deposits and withdrawals | none read yet | 2,381,932 | 728,657 | 79,258 |
| Venues holding it | — | 9Morpho Vaults V2 54% | 22Morpho Vaults V2 47% | 4Lista Lending 98% |
An outside reader put it exactly right, and the page used to deserve it: we are counting whatever rail we have wired on each network, and those rails are not the same thing. A chain with an ERC-8004 registry and nothing else will show hundreds of thousands of identity events and no work at all. That is a fact about the registry and about our reading, not a ranking of agent economies. The two are counted under separate headings here for that reason, and a total over both would be a number with no unit.
Even inside one heading the nouns differ: a job, a task, a payment, a channel and an escrow are five things, and they are added under “work” because they are all somebody doing something for somebody, not because they are interchangeable. Each rail’s own unit is printed beside its name.
And a network with no capital column is a network we track no vaults on, which is a fact about this site. Everything here is bounded by what we read; each network page names what it leaves out.