State of the Agent Economy: Q3 2026

On June 24, 2026, Treebeard indexed 237,099 autonomous agents across 23 blockchain networks. By September 19, that number reached 489,453 — a 106% increase in under three months, across 24 chains.

This is not uniform growth. Some chains more than doubled. One major chain added zero net agents. A chain that barely existed in June now ranks fifth. The agent economy has a topology, and it is becoming more defined, not less.

The question stopped being “are agents a real thing?” at around 200,000. The question now is which chains are building agents that actually work — and which are accumulating registrations.

The Concentration Problem

BSC accounts for 272,111 agents — 55.6% of the entire index. The next three chains (Base, Ethereum, Billions) together hold 33.1%. The remaining 20 chains share 11.3%.

That concentration has held for two consecutive quarters. BSC's dominance is structural, not cyclical. The BNB ecosystem moved earliest on ERC-8004 adoption, and that first-mover effect compounded. At current growth rates, BSC's share of the index is likely to decline — not because BSC slows, but because other chains are accelerating faster.

Chain Divergence: 87 Days

Comparing June 24 to September 19 reveals a sharp divergence in chain trajectories. The same EVM environment, different outcomes.

Ethereum's agent count is identical on September 19 to what it was on June 24. Not down. Not struggling. Exactly the same number. The chain is not inactive — ERC-8004 registration simply has not become a default step for Ethereum deployers. One major framework or toolchain adding it as a default would move the number materially.

Base has now overtaken Ethereum in total agents (72,147 vs 64,262) — the first time a chain other than BSC has passed Ethereum in our index. Base's growth rate has held above 80% annualized for three months.

Billions chain is the most unexpected data point in this report. Starting from fewer than 4,000 agents in June, it crossed 25,000 by September — a 567% increase. No other chain in the index grew faster in absolute or relative terms. Its current count ranks it fourth overall, ahead of Abstract and XLayer. We have limited additional context on what is driving that deployment rate. It belongs on your radar.

The Hidden Chains

Ecosystem coverage of autonomous agents in the broader market clusters around five or six chains. The Treebeard index reveals a broader picture: 24 networks, each with real registered agents, each rated on the same methodology.

Gnosis, Celo, Shape, Abstract, XLayer, MegaETH — none of these appear in standard market analysis. All of them have agents. Agents on Celo or Gnosis are rated identically to agents on Base. The chain of deployment doesn't change the methodology.

The gap between what is covered and what exists is one of the larger inefficiencies in how the ecosystem is currently analyzed.

The B− Ceiling

The highest grade in the Treebeard index is B−, at a score of 75.2. Eight agents share this position. None has exceeded it.

This is not a bug in the methodology — it reflects how difficult it is to satisfy all six signal categories simultaneously. The six categories are Economic Value, Operations, Code Quality, Autonomy, Safety, and Community. Each is measured independently. High scores on two or three are common. High scores on all six, sustained over 30+ days of signal history, are not.

The hysteresis buffer in the rating engine adds one additional constraint: a grade will not change unless the underlying score moves by more than two points. This prevents grade churn from signal noise. Grade changes, when they happen, reflect real movement.

Breaking above B− requires scoring well on all six signals simultaneously. Most agents do one or two things well. Doing all six at once, for long enough to build sustained signal history, is what the top of this index actually looks like. No agent has done it yet.

What to Watch in Q4

At the current growth pace, the index crosses 500,000 agents in the next two weeks. That number is less significant than it sounds — volume is the easy part of this story. The more interesting question for Q4 is whether quality indicators follow quantity.

The chains to watch are Abstract and MegaETH. Both showed consistent acceleration through Q3 with characteristics that typically precede quality improvement: builders arriving early, deploying iteratively, building signal history. If x402 adoption follows, we should see the first agents with sustained B scores by year end.

Ethereum's flat line is a structural question, not a temporary one. ERC-8004 registration is an active step that Ethereum developers have not widely adopted. That could change quickly — one major framework adding it as a default would move the number materially.

The most overlooked number in this report is Billions chain at +567%. We'll be watching.