THE AGENT ALMANAC
Vol. I · No. 1 · Free public reference for the agent internet
The scoreboard · settlement-layer truth

The Agent Economy, measured on-chain.

Not listing counts, not press releases — actual stablecoin settled to the wallets that agent services get paid at. We count every payment we can verify at the settlement layer and tell you exactly what we can’t yet see.

Snapshot 2026-09-05. Tracking since 2026-07-11 (58 daily points).

Validated settled volume
$3,885,842.47
Excludes single-payer self-funding & test traffic. Raw on-chain: $4,534,114.09.
Real providers
2,095
Wallets that have earned > $0. Not the 72,159 listings.
Distinct payers
38,834
Unique paying addresses seen at the settlement layer.
Measurement coverage
80.7%
2,607 of 3,231 listed wallets are on chains we measure today.

Read this before you quote the number above

Step 1 · the whole measured economy
$4,534,114.09

Every stablecoin payment we can verify at the settlement layer, wash included.

Step 2 · it is not evenly spread
74.6%

of all settlement goes to a single counterparty (Bitrefill). Most of what looks like an economy is one participant.

Step 3 · what is left
56.4%

of the remaining $1,149,685.84 — agents actually paying agents — is wash by our published rules.

We lead with this rather than bury it. The concentration is visible in our own Gini (0.996275) and top-wallet share, so anyone checking would find it in a minute; a number quoted without it is misleading even when it is arithmetically correct. The step-2 figure is pure arithmetic and needs no interpretation. We additionally classify Bitrefill as consumer commerce rather than agent-to-agent settlement — a documented judgement, not a heuristic, recorded with its reasoning and versioned by wash_rule_version. That classification is supporting detail: step 2 holds whether or not you accept it. Both views are sealed into the integrity chain, so neither can be quietly revised later.

Read the number right

Directories advertise 72,159 listings. But only 2,095 wallets have ever been paid — about one real earner for every 34 listings. We lead with what got paid, and treat the listing count as context, never the headline.

Raw is not real

Of $4,534,114.09 in gross on-chain settlement, $648,271.62 (14.3%) across 171 wallets is single-payer self-funding or non-production traffic — e.g. one wallet takes tens of thousands from a single payer. We report the validated floor of $3,885,842.47 and show the raw number beside it, never instead of it. How we adjust →

Where the money settles

ChainSettled volumeShareReal providersActive (24h)
base $1,650,377.80 36.4% 1,765 440
arbitrum $1,299,502.85 28.7% 12 4
polygon $1,242,271.64 27.4% 20 3

Base, Polygon, Arbitrum, Ethereum and Solana measured directly via on-chain USDC/USDT/PYUSD Transfer logs. Other non-EVM rails (Stellar, Noble/Cosmos) are not yet in these totals — see coverage below.

How concentrated is it?

Top wallet’s share
74.6%
of all measured settled volume.
Top 10 wallets
92.9%
of all measured settled volume.
Volume Gini
1.00
0 = evenly shared, 1 = one wallet takes all.

Today the economy is heavily concentrated — a single wallet accounts for 74.6% of measured volume. Early markets look like this; we’ll track whether it decentralizes over time.

Listed price vs. what actually gets paid

Median quoted price
$0.01
Across live 402 quotes, 7-day window.
Avg. paid per transaction
$0.32
Measured settled volume ÷ transactions.

Different populations (all quoted services vs. wallets that actually earned), shown side by side so you can judge whether listed prices reflect real demand.

Reality check

Endpoints that respond
94.0%
392 of the 417 endpoints we have actually contacted. That is 0.6% of the 70,167 catalogued, so this is the response rate among endpoints we tested and NOT the share of the agent economy that is alive. Measured 2026-09-06. Corrected 2026-08-03: this box previously divided by the whole catalogue and read 0.3%.
Payments under 10¢
95.6%
The agent economy is a micropayment swarm.
Payer retention
61.8%
770 returning of today’s payers.

Payment-size distribution — 42,195 transfers

<$0.01
79.2%
$0.01-0.10
16.4%
$0.10-1
1.8%
$1-10
1.3%
$10-100
1.0%
$100+
0.4%

Asset mix: USDC 99.6% · USDT 0.3% · USDC.e 0.0% · 0x0b2c639c 0.0%

Uptime, size distribution, asset mix and payer retention are captured forward-only from on-chain settlement and live endpoint probes.

What these numbers do & don’t include

Full details on how we measure →

Operate a service and see something wrong? Email hello@agentalmanac.org — we’ll fix it.

The other half — reported, not measured

Everything above is measured on-chain. But most of the world’s agent payments happen off-chain, inside Asian super-apps and card networks — walls we can’t see behind. We track what they report so you get the whole board; these figures are operator-reported, not verified by us, and never mixed into our measured numbers.

OperatorRegionProtocolReportedAs of
JD.comChinaA2P2 (Agent Autonomous Payment Protocol)Published autonomous-spending security standard2026-06
MastercardGlobalAgent Pay for MachinesMachine payments across cards, accounts & stablecoins; live SG/MY2026-06
GrabSE AsiaGrab Intelligence Layer13 AI-powered agentic products2026-04
Ant Group / AlipayChinaAI Pay + Agentic Commerce Trust Protocol~120M AI-agent transactions in one week; 100M+ AI Pay users2026-02
Tencent / WeChatChinaWeChat Pay agent (WorkBuddy, QClaw)Small-scale AI-payment tests inside 1.6B-user platform2026
UnionPayChinaAPOP (Agent Payment Open Protocol)Standard for agent-initiated payments2026
Ant InternationalCross-borderAMP + Antom stablecoin-for-merchantsOpen agent-payment framework; USDC merchant acceptance; Alipay+ ~2B users2026

Reported by the named operators via public announcements. The on-chain economy we measure is a fraction of this; the two rails may converge at cross-border stablecoin settlement.