⌘K

Live evidencecounted 3m ago

0.48days of MEME volume to unwind the crowd99%of the XMR book on one side, across 28 vaults$1.30Bunder one key with no guardian, 52 of the 424 vaults whose control we have read$334.1Mwhose control we have never read, across 3,229 vaults$1.95Bsitting in vaults and not lent out$11.27Bunder 3,653 vaults we read$4.79Bin 323 vaults where we saw automated execution, nothing declared2,080vaults declared or on a venue built to run software, holding $1.01M324software-run vaults scored, 13% of them450operators identified behind them26networks where we read capital, from Ethereum to Cardano31,298,687deposits and withdrawals read from chain5changes recorded in the last day588,846ERC-8004 agent registrations read−0.61/ 30 days360,982distinct owners behind those registrations199,780registrations carrying a readable card, 34%3,330,722paid requests on the Olas rail (deliveries are counted apart)−0.69/ 30 days87suppliers delivering that work411,027agent-to-agent jobs read on Base+0.04/ 30 days$1.0Msettled between agents, all time1,175suppliers listed on the ACP rail189agents live on Cardano76Cardano agents deregistered, 29% of all ever minted29,141escrows between Cardano agents since Jun 2025, 142 still open78%of every agent payment we read comes from 2,571 addresses that never stop307,376addresses paid once and never came back, of 494,774 that ever paid934of those payers are an agent we can name, 0.189% of them635agent hosts answered when we asked, of 2,532 probed17,125agents sit behind a host that answers nothing at all404msmedian answer from the 635 hosts that answer44%of all agent ratings were written by ten addresses, of 671,178 ratings read29,502rated agents have exactly one rater, of 47,110 rated at all86,979service endpoints declared across 33,840 agent cards we read0.48days of MEME volume to unwind the crowd99%of the XMR book on one side, across 28 vaults$1.30Bunder one key with no guardian, 52 of the 424 vaults whose control we have read$334.1Mwhose control we have never read, across 3,229 vaults$1.95Bsitting in vaults and not lent out$11.27Bunder 3,653 vaults we read$4.79Bin 323 vaults where we saw automated execution, nothing declared2,080vaults declared or on a venue built to run software, holding $1.01M324software-run vaults scored, 13% of them450operators identified behind them26networks where we read capital, from Ethereum to Cardano31,298,687deposits and withdrawals read from chain5changes recorded in the last day588,846ERC-8004 agent registrations read−0.61/ 30 days360,982distinct owners behind those registrations199,780registrations carrying a readable card, 34%3,330,722paid requests on the Olas rail (deliveries are counted apart)−0.69/ 30 days87suppliers delivering that work411,027agent-to-agent jobs read on Base+0.04/ 30 days$1.0Msettled between agents, all time1,175suppliers listed on the ACP rail189agents live on Cardano76Cardano agents deregistered, 29% of all ever minted29,141escrows between Cardano agents since Jun 2025, 142 still open78%of every agent payment we read comes from 2,571 addresses that never stop307,376addresses paid once and never came back, of 494,774 that ever paid934of those payers are an agent we can name, 0.189% of them635agent hosts answered when we asked, of 2,532 probed17,125agents sit behind a host that answers nothing at all404msmedian answer from the 635 hosts that answer44%of all agent ratings were written by ten addresses, of 671,178 ratings read29,502rated agents have exactly one rater, of 47,110 rated at all86,979service endpoints declared across 33,840 agent cards we read

Status and service level

What the API is committed to, and what it has actually done. Both are here because a target without a measurement is a sentence, and a measurement without a target is a number nobody can hold us to.

Last request served 1 minute ago

This is an error rate, not an uptime measurement

Every figure here is computed from requests that reached the API. A minute nobody called writes nothing, so a quiet night and an outage look identical on this page. An outage in which requests never arrived at all, because the proxy or the certificate failed, cannot lower any number below. Read it as how often we answered badly when asked, not as how often we were there to ask. A per-minute probe that would close the gap is not built yet, and this notice stays until it is.

Requests answered well, 30 days

99.777%

of 173,391 calls that reached us · target 99.5%

Requests answered, 24 hours

99.818%

14 of 7,698 calls that reached us failed

Median response

≤ 25 ms

over 30 days · a bucket's upper edge, not an exact figure

95th percentile

≤ 1,000 ms

target under 800 ms · a bucket's upper edge

How long a request takes

Counts rather than an average, because a mean hides the tail completely: a hundred fast calls and one that took eight seconds average out to something healthy, and the eight seconds is the one you noticed. Each row is its own band, not a running total: the shares add up to 100%. The percentiles above are read off these bands, so they are the upper edge of the band the percentile falls in, never an exact millisecond.

0–25 ms53.9%

25–50 ms15.4%

50–100 ms11.9%

100–250 ms10.2%

250–500 ms3.4%

500–1,000 ms1.5%

1,000–2,500 ms1.8%

2,500–5,000 ms0.8%

over 5,000 ms1.0%

Whether every tool answers

21 of 21 answered when last asked. None was silent.

Tool	Answered	Took	Last asked
agent_changes	yes	14 ms	Read 3 minutes ago
check_service_endpoint	yes	9 ms	Read 3 minutes ago
counterparty_bulk	yes, with its price	159 ms	Read 3 minutes ago
counterparty_check	yes	5 ms	Read 3 minutes ago
counterparty_history	yes	6 ms	Read 3 minutes ago
evaluate_action	yes	7 ms

What we commit to

99.5% of API calls answered: Measured over a calendar month, counting only failures that are ours. A 402 is the price working, a 404 is a request for something absent, and a 429 is a quota doing its job, none of those is an outage, and counting them would let a mistyped slug read as one while hiding a real failure inside the noise.
95% of calls under 800 ms: Server time, from request to response. It excludes the network between us and you, which we cannot measure and will not claim to.
Notice before anything breaks: A removed field or a changed meaning gets thirty days’ notice by email to every account with a key. Adding a field is not a breaking change and will happen without warning.
If a month misses: a month of the plan, credited without being asked for: Applied from this page’s own numbers rather than on request. Somebody who has to notice and complain to be compensated is somebody being charged for our failure to notice.

The limits of this page

This is one machine in one place. There is no failover, so a hardware fault or a network partition at the host is an outage, and the honest expectation is hours rather than seconds. The measurement is taken inside the server: a request that never reached us cannot appear here, which is precisely the failure you are most likely to experience and least likely to see reflected. Planned maintenance is announced in advance and counted against the number like any other downtime, an availability figure with the bad days excluded is not an availability figure.

Something broken? How to reach us.