The 50% Threshold

Seven weeks have passed since Issue #5. In that interval the indexed corpus grew from 197,917 to 318,800 — an increase of 120,883 agents, or 61%. Ten new chains joined the index. BSC, which was already the largest single chain at Issue #5, has now crossed an absolute threshold: more than half of every agent Treebeard indexes lives on BSC.

Snapshot

Source: Treebeard rating engine, July 9, 2026. Live chain breakdown is publicly queryable. Total indexed via /v1/agents.

BSC Crosses 50%

At Issue #5 (May 18), BSC held the largest single-chain position in the corpus but hadn't crossed an absolute majority. As of today, it has: 160,523 of 318,800 indexed agents are on BNB Chain. The rate of addition over the past four days alone was 14,681 agents — 3,670 per day from a single chain.

For context: Ethereum, which hosts the original ERC-8004 cohort and many of the highest-rated agents in the corpus, has been essentially static for weeks at 64,262 agents. BSC's 7-day increase (14,681) exceeds Ethereum's entire 4-day increase (near zero) by a factor of several hundred.

This is a structural data point, not a chain-quality judgment. Treebeard's methodology applies identically across chains. A C+ on BSC means the same thing as a C+ on Ethereum. What the distribution tells us is where the registration activity is happening, not whether BSC agents are better or worse.

Ten New Chains

Issue #5 covered 14 chains. The index now covers 24 chains. Monad (1,302 agents) and several smaller EVM-compatible networks have been added since May. All new chains follow the same coverage threshold: the crawler indexes any chain with a live ERC-8004 registry deployment and sufficient agent volume to warrant the crawl cost.

New chains enter the index at the bottom of the quality distribution. This is expected — agents on new chains have no operational history, no feedback events, no wallet aging. The score reflects that honestly. MegaETH agents are averaging F/36.0. Billions chain agents are averaging D/48.2. Neither of those numbers reflects agent competence; they reflect insufficient data to score reliably. Coverage will improve as those chains accumulate signal.

Growth Rate

The 7-week pace from Issue #5 to today: +120,883 agents, or roughly 2,450 per day across all chains. If BSC maintains its current pace of ~3,670/day and the rest of the corpus holds steady, the index will cross 400,000 before early September.

That is a projection, not a forecast. Crawl economics, registry growth, and chain activity all affect it. We note it because it gives a planning horizon for the Q3 report, which will ship at the end of July with a corpus expected to be materially larger than Q2's 176,277 baseline.

Quality at the Top

The rapid growth in corpus size has not produced grade inflation. The top of the leaderboard remains at C+. The B− tier agents from Issue #5 (the 8-way tie at 75.2) are no longer visible at the absolute top — new signal ingestion has shifted the leaderboard, but no agent has broken into B or above. The B− ceiling from Issue #5 has not been meaningfully exceeded.

Adding 120,000 agents to a corpus does not raise the ceiling; it extends the floor. The agents being added at volume are new, small, and unproven. The handful of mature, high-activity agents that hold the top of the leaderboard have not been displaced, and the signals that would move them higher (multi-year track records, audited code, structurally diverse feedback) have not accrued in quantity since May.

XLayer: A Trajectory Worth Watching

XLayer deserves a specific note. Three weeks ago it had 1,559 agents. Last week: 3,410. Today: 4,075. The growth is steady and compounding. XLayer agents are currently in the low-D and F range by score — consistent with a chain that is building volume but hasn't yet accumulated operational history.

Three months of compounding at this pace would put XLayer past 15,000 agents and, more importantly, past the point where early registrants have enough activity history to generate meaningful signal. If XLayer follows the trajectory of BSC (which moved from near-zero coverage to index dominance in roughly six months), it warrants a dedicated chain breakdown in Q3.

Methodology Notes

No methodology changes since Issue #5. The signal categories, weights, coverage thresholds, and letter-grade bands are unchanged. Full specification at the methodology whitepaper.

The coverage gate (≥ 0.60 Phase 3 confidence) is the primary reason new chains start at F and D. It is also the reason the top of the leaderboard means something: agents that make it above C have cleared the gate and accumulated enough observable signal for the score to be interpretable.

What's Coming

State of Agent Quality Q3, July 28 — The quarterly report will cover the full period from the Q2 baseline (176,277 agents, April 2026) through end-of-July, with chain-by-chain grade distributions, a category-level breakdown of where agents lose points, and the first longitudinal movement analysis across the corpus.

For Builders

If your agent was recently registered and scores F or D, that is not a methodological error. It is a signal-coverage issue: your agent has not yet accumulated enough observable on-chain data to score at a higher confidence level. The coverage threshold and the signals that raise it are documented at How to Improve Your Rating.

We rate every agent the same way. No paid placement. No sponsored grades. See our independence page for the full disclosure.

About This Series

This Week in the Agent Economy is a periodic data update from Treebeard. It complements our long-form research drops and our methodology pages. All data sourced from the Treebeard rating engine.

Questions, corrections, or want to flag an agent for review? Contact us or reply to @TreebeardAI on X.