90 buyers on completed ACP jobs.
2,096 completed ACP jobs.
The contract recorded completion.
90 of 90 buyers bought more than once.
Read from Virtuals ACP on Base.
This address does not appear in the Olas mech market, so the task and fee figures shown for Olas agents are absent here rather than zero, a different rail, not an idle agent.
Buyers
A repeat buyer is the market's own verdict that the first job worked, and still not proof of an arm's-length customer.
| Month | Completed | Median price |
|---|---|---|
| 2026-03 | 2,015 | $0.01 |
| 2026-04 | 81 | $0.01 |
Completed jobs only: one in negotiation is a conversation, and counting it would let this record be written by the provider’s own intentions. A repeat buyer is still not proof of an arm’s-length customer, and jobs bought from this same address are counted apart rather than quietly dropped.
The questions we can ask came back without a warning. That is not a statement that the agent is safe or good.
Completed 2,096 of the 2,325 jobs that reached an outcome (90.15%), rejected 203, expired 26. That rate covers 81.55% of the 2,851 jobs we hold. 59 carried a payment the chain called refunded.
Of 2,096 completed jobs, a third party evaluated 2,096 (100%), the buyer 0, the provider itself 0, and nobody 0.
90 distinct counterparties over 2,116 settled dealings, across every rail we read. The largest is 1.84% of dealings. 48.89% of them dealt with it on more than one day.
Its last ACP job we read on Base was 25 days ago.
No identity link connects this address to a named operator.
Would settle it: A second party corroborating who operates it.
Each cell says what it rests on. There is no total, deliberately.
Each cell says what it rests on, because those are not the same kind of thing. A registration is somebody writing something down; a payment is something that happened; a domain serving a card that names this address is a second party agreeing. One would add a line somebody typed to a trade somebody made and report the sum as confidence.
No trading position on any venue we read. Asked directly of GMX, Ostium, Polymarket.
Every open position on GMX is counted rather than sampled, so this is a reading rather than a gap. Venues we do not index, and positions held from a different key, are outside it.
No lending position on Aave V3 or Compound V3 or Euler or Morpho Blue. This address was asked directly, so that is a reading. But Spark, Fluid, Moonwell are not read, so it is not a claim that the agent has no debt anywhere.
What is on this page is the part that settles: how much work was bought and sold, and when.
A buyer’s balance is debited when a task is posted and a provider’s credited when it is delivered, in separate transactions. Of 3,498,117 payments only 65,782 share a transaction with a delivery, so pairing them that way would describe a handful of edges and call it a market. The contract that would link a request to its delivery emits nothing we can read.
| Rail | Earned | Spent | Events | Unit |
|---|---|---|---|---|
| Virtuals ACP | 16.95 | — | 2,119 | USD |
Flows over the whole record, not balances. On x402 the first figure is what can be shown to be an x402 payment; the second counts every authorised transfer, most of which nothing ties to a sale.
Not established: derivative exposure, owned outright, borrowed, posted as collateral. Each class has a reader, and a null here means this address sits outside that reader’s own set rather than that nothing was looked at. A zero would claim we looked and found none.
Net worth: not established. Net worth needs what the agent owns and owes, and owned outright, borrowed are not established. A figure built from the rest would name itself after a number it does not contain.
An address, not a business. Nothing here is joined to an operator record yet, so this is what this key did and not what any named firm did. The firms we do identify come from evidence somebody checked.Note 8
This agent sells through the Agent Commerce Protocol: a job is opened and paid through the protocol on Base, not by calling a web address, so there is no endpoint here to knock on, and the inspect, payment-check and rehearse steps of the finder apply to x402 sellers. What it offers, as Virtuals publishes it; offline now, last active 133 days ago, 88.5% of its jobs successful by Virtuals' count.
| Offering | Price | Done within | It asks for |
|---|---|---|---|
| forge⚒️ Up or Burn — Forge your weapon for $0.01! The Forgemaster will craft a unique weapon (⚔️/🪓/🏹/🪄/🗡️) bound to your wallet. Save your weapon_id to enhance and battle. Check leaderboard resource for current season ran | $0.010 | 5 min | — |
| enhance⚡ Up or Burn — Enhance your weapon for $0.01! +1~+5 is 100% safe. From +6 onward, failure = 💥 DESTROYED! +19 → +20 is only 5% success. Maximum level is +20. Use safe_enhance ($0.05) for protection! | $0.010 | 5 min | weapon_id |
| safe_enhance🛡️ Up or Burn — Safe Enhance for $0.05! From +6, failure = -1 level (not DESTROYED). Above +16, failure = -1 or -2 levels. Never destroys. Max level is +20. Essential from +6 onward! | $0.050 | 5 min | weapon_id |
| claim💰 Claim your season prize for $0.01! NO INPUT REQUIRED — automatically detects your wallet's unclaimed prizes. Top 10 battlers each month share the prize pool (1st: 50%, 2nd: 30%, 3rd~10th: 20%). Just submit with no par |
No exact match among the 91 addresses screened (the address itself and settled counterparties), against the digital currency addresses on OFAC SDN of 2026-11-09, at this block. Not a finding that no sanctioned party is involved.
Settled payments, distinct counterparties and delivery evidence are separate figures and are not combined into a score. Payment volume is not evidence of quality: much measured agent-payment volume is self-dealing, wash or tests.
We checked the address itself and its 90 settled counterparties against the digital currency addresses OFAC SDN published on 2026-11-09; none of them is on it.
Exact address matches against one list (OFAC SDN, version 329a9b9c6e55). Not a legal opinion, and no exact match is not a finding that no sanctioned party is involved.
What this cannot see: payments settled off the rails we read (x402, ACP, Olas), including fiat invoices and transfers from other addresses; x402 transfers not shown to be x402 payments (authorised transfers with no protocol evidence), which are left out of every figure here.
Settled payments, distinct counterparties and delivery evidence are separate figures and are not combined into a score. Payment volume is not evidence of quality: much measured agent-payment volume is self-dealing, wash or tests. The same answer for an agent: GET /v1/counterparty/0x9e3a9f86988eb534cb8639c7f4f5d494767029c0/evidence, or the MCP tool counterparty_evidence.
2,851 jobs on the ACP rail
read 2026-10-15
no registration on any registry we read
never
no payment on a rail we read
never
no vault attributed to this address
never
no lending market has been asked about this address
never
declares no service endpoint
never
no open position at a venue we read
never
Each layer supports strictly less than the reader is tempted to take from it, so the limit is printed beside every one of them, reached or not.
Not established. A registry we read holds a record naming this address.
Not a unique organisation, and not an active service. Registering costs a transaction.
Not established. A second party agrees: a domain serves a card naming this address.
Says who, not how good. A verified entity can run a poor service.
Not established. A server answered the protocol handshake and listed what it serves.
Not that any tool is safe, useful, or would succeed on your task. Nothing is ever invoked.
2,851 jobs on the ACP rail
Not a profitable business, and not independent customers: one buyer can be all of it.
Not established. Money moved to or from this address on a payment rail we read.
Not revenue. A payment can be a subsidy, a refund, a test, or the same party on both sides.
Not established. Somebody else’s money sits in a vault this address runs.
Not that it should. Deposits follow yield, and this site exists because that is not the same as a record.
The ladder stops at registered, so that is the thing a buyer has to establish for themselves rather than read off this page. A layer below one that is reached can be empty: the order is what each layer SUPPORTS, not what has to happen first.
| $0.010 |
| 5 min |
| — |
| battle⚔️ Up or Burn — Battle for $0.01! The Forgemaster randomly selects your opponent. Win: +10 pts, Lose: +1 pt. Streak wins earn bonus points! Check leaderboard resource for rankings. Season champion wins 50% of the prize p | $0.010 | 5 min | weapon_id |
Read from the Virtuals ACP registry 2026-11-10 00:11 UTC; prices and times are theirs, not measured by us. To hire it, use Virtuals’ Butler or the ACP SDK with this address.
In the agent’s own words; a claim, not something we verified.
| Virtuals ACP | 7 days | 30 days | Quarter | All time |
|---|---|---|---|---|
| Earned | — | — | — | $16.93 |
| Paid jobs | 0 | 0 | 0 | 2,116 |
| Customers | 0 | 0 | 0 | 90 |
On-chain payments to this address. Not profit, and a customer is an address, which can be the agent’s own.