What agent-managed capital is actually doing, read from public on-chain and venue data, with the gaps named rather than filled in.Note 2Note 3Note 5
Exposure
Every open vault we read, agent-managed or not. A chain absent here is one we do not observe, not one with nothing on it.
Exit capacity
Cash plus lent plus drained equals the $8.63B supplied, with no dollar in two places. Read 2026-10-16 from each register's own supplied and borrowed figures, not against stated size, those are written by different jobs at different times.
From $9.49B to $10.79B. A line going up reads as a market growing, and three different things move it.
An arrival is not growth: some of these vaults are new, and some existed all along until we began reading them. The two cannot be told apart from this side, so neither is claimed. What is not attributed to flows is not called performance. Flows are read where the venue publishes them; for the rest the change stays unexplained, because calling an unread deposit a return is the oldest bug on this platform.
It begins on 2026-08-17, the first day every platform shown here was being priced, so every day of it compares like with like. Morpho Vaults V2 and Hyperliquid have value history back to 2026-02-13, on their own vault pages; stacked here it would be one venue's record against the others' absence. Vaults that closed before we read them are absent, so the early days are a floor.
read 2026-10-16
Two ends of the same reading: money that cannot come out, and money that is not being put to work. Every input is public, the cross-vault view is ours.
Depositors here are waiting on borrowers rather than on a market: the capital is committed, not gone. Utilisation answers both questions and only the first is usually published.
Read the total as an upper bound. $6.89B, 61%, is out with borrowers and could be counted twice. 40 vaults hold another tracked here, putting $217.92M in the total twice.Note 1