THE AGENT ALMANAC
Vol. I · No. 1 · Free public reference for the agent internet
Archive ·

Issue No. 1 — The 90-minute audit

The Agent Almanac · Vol. I · No. 1 · April 2026


We pulled 90 minutes of x402 activity off Coinbase's own Agentic Market feed today. Here is what we found.

Of 300 transactions we observed:

Of 87 unique buyer wallets:

Of 28 unique recipient wallets:

The full economic activity in our 90-minute slice was about $85, almost all of it from a handful of $7-$12 outlier payments. The rest — the 91.5% of transactions that were exactly $0.001 — was dust.


What this means for the headline numbers

Coinbase's Agentic Market dashboard (the official one) shows 48,912 unique buyer wallets and 1,559 unique seller wallets with $1.12M cumulative x402 volume since the protocol launched.

Press coverage has been citing $50M lifetime, sometimes $600M.

After watching 90 minutes of activity, our honest read on what those numbers actually represent:

  1. "48,912 unique buyers" is technically true — that many wallet addresses have signed at least one x402 payment. But our sample suggests roughly 77% of "buyers" are one-shot wallets, most likely test/demo activity from AgentKit, dev tutorials, and short-lived agent sessions. Each demo creates a fresh wallet.

  2. "1,559 unique sellers" includes a long tail of ghost services. In our slice, more than half of recipient wallets received exactly one transaction. The dollar volume concentrates into a handful of real services.

  3. The $1.12M lifetime number, divided by 803,968 transactions, gives an average tx of $1.40 — but the median is $0.001. The mean is being pulled by a small set of larger payments while the bulk is sub-cent.

  4. The "$50M-$600M" press numbers don't match Coinbase's own dashboard. They likely include test traffic, multi-chain x402 facilitator activity outside the curated Agentic Market, and possibly aggregations that double-count routing hops.

  5. Artemis published an estimate that ~50% of on-chain x402 activity is wash-traded or self-dealt. Our 90-minute sample is consistent with that — the dust patterns we saw look more like automated loops than agent-to-merchant commerce.

What "real" agent commerce looks like (in our 90 minutes)

If we strip $0.001 floor payments and look only at what's left:

That last number is what we believe the real x402 economy is doing right now. Not zero. Not $50M lifetime. About a half-million-dollar annual run rate, growing.

That's a real number worth tracking. It's also a much smaller number than the press has been printing.

We built wallet profiles to make this checkable

Every wallet that appears in our scraped activity now has its own profile page. You can see classifications (one-shot tester / wash-suspect / active service / etc.), top counterparties, dollar volume, and a Basescan link to verify on-chain. Try one:

https://agentalmanac.org/wallets/0xa9dd7cc9cbf0e05551332209289f04be36bc2315

That's the "mega" recipient — 187 sub-cent transactions to 19 cents of revenue. Classified as "dust accumulator." Decide for yourself whether you'd call that real commerce.

What we're going to do next

Read the dashboard

agentalmanac.org — live now. Free. No signup. Public API at /api/*. Agent-readable manifest at /llms.txt. Per-wallet profile pages at /wallets/{address}.

Email hello@agentalmanac.org with corrections.

The editor


The Agent Almanac is published from upstate New York by an independent operator. We do not accept funding from agent-economy participants. Free for the audience, monetized by those who serve them.

← All issues Subscribe by email →