In the seven days to November 13, 2026, 50.3% of the 34,396 new ERC-8004 agent registrations came from a single source (centralcity.ai); without it, 17,100 remain. On x402, the largest facilitator (Polygon's) settled 62.9% of 1,774,065 transactions over the 5 complete days to November 13, 2026. These are measured shares, recomputed daily, and they say how much of each headline number depends on one party.
Daily ERC-8004 registrations excluding each day's largest source · agents/day
Daily ERC-8004 registrations excluding each day's largest source, in agents/day
Period
agents/day
09-15
822
09-16
1,902
09-17
1,218
09-18
1,254
09-19
581
09-20
903
09-21
637
09-22
1,214
09-23
805
09-24
652
09-25
1,153
09-26
1,124
09-27
1,510
09-28
1,208
09-29
1,007
09-30
2,014
10-01
498
10-02
1,135
10-03
621
10-04
549
10-05
824
10-06
1,495
10-07
877
10-08
965
10-09
1,629
10-10
744
10-11
1,241
10-12
1,371
10-13
3,228
10-14
1,104
10-15
1,269
10-16
971
10-17
872
10-18
704
10-19
809
10-20
1,235
10-21
2,852
10-22
1,063
10-23
13,162
10-24
2,638
10-25
6,472
10-26
2,497
10-27
2,521
10-28
2,603
10-29
1,958
10-30
2,024
10-31
1,594
11-01
2,046
11-02
2,373
11-03
2,069
11-04
3,583
11-05
6,993
11-06
4,279
11-07
2,131
11-08
1,641
11-09
2,473
11-10
2,509
11-11
4,258
11-12
1,026
11-13
1,172
The rule
Headline counts in this field are cheap to move. Registering an agent is one transaction; a payment rail can be driven by one integration. So next to each total that can be split by source, this page reports two more figures: the share that comes from the largest single source, and the total with that source removed.
The rule is mechanical and the same everywhere: rank the sources by count, take the top one, subtract it. It does not decide whether that source is legitimate, automated, or valuable — a large platform onboarding real users and a script minting identities look the same to it. What it tells you is how much of the number depends on one party, which is the question to ask before reading a total as broad adoption.
What a source is in each series
ERC-8004: the host of the agent-card URL written into each Registered event, across every chain the pipeline scans, testnets excluded. Registrations sharing a host come from one application, whatever the number of owner addresses behind them.
x402: the facilitator that submitted the settlement, attributed through this site's reviewed address registry. The window is the set of complete days carried in the live feed, which is shorter than a week.
Virtuals ACP: jobs on Virtuals' current contract on EVM chains. There is no single-source split here yet, so the page reports how many jobs complete and how many distinct buyers and sellers appear instead. ACP on Solana is not covered.
What is not split yet
Olas, Tempo MPP, Masumi and Base's agentic ecosystem are reported as totals only. Their figures are not adjusted on this page, and no concentration claim is made about them in either direction.
Also asked
Does a high concentration mean the activity is fake?
No. It means the number depends on one party. A single application onboarding many real users and a single script creating many identities both produce a high share. The figure tells you how much weight a headline total can bear, not what the activity is worth.
Cite this page
agent economy, “How concentrated is the agent economy?”, agenteconomy.to/stats/how-concentrated-is-the-agent-economy, figures as of November 13, 2026 (UTC). Recomputable from agenteconomy.to/data.json.