Skip to content
Prepaid, no subscription trap. Top up a balance and we draw the monthly fee from it.
tcggraph

Status

Measured, not promised

Every figure on this page is computed from the API's own request log and price tables, over the last 30 days, and recomputed every hour. The method for each is written out below it.

96k
requests served
Since Sep 8, 2026, 4:04 PM UTC, by 12 accounts.
73 ms
median response time
p95 279 ms. Server-side, every request, cached or not.
99.998%
responses without a server error
2 responses with a 5xx status in the window.
9 hours ago
Cardmarket last read
Each source's most recent successful read is listed below.

Computed Sep 17, 2026, 11:57 AM UTC. Figures are recomputed at most every 60 minutes.

By day

The last thirty days

Requests, median latency and server errors per UTC day. A quiet day is a quiet day, not a gap in the log.

DayRequestsVolumep505xx
Sep 178,870
81 ms0
Sep 1623,793
71 ms0
Sep 1521,550
76 ms0
Sep 1419,961
76 ms0
Sep 1317,877
68 ms0
Sep 122,815
70 ms0
Sep 11360
59 ms0
Sep 10550
82 ms2
Sep 9465
57 ms0
Sep 8171
93 ms0

Price sources

When each marketplace was last read

A quote's own updatedAt field carries this per card. Here it is per source, so a stalled feed shows up before a customer notices.

Cardmarket
9 hours ago
Sep 17, 2026, 2:55 AM UTC
TCGplayer
6 hours ago
Sep 17, 2026, 5:33 AM UTC
Card Kingdom
6 hours ago
Sep 17, 2026, 5:33 AM UTC
Manapool
6 hours ago
Sep 17, 2026, 5:33 AM UTC

Method

How every number is produced

Requests served

A count of rows in the request log for the window. Every authenticated call to the REST or GraphQL API writes one row after the response is sent, whatever the status code and whether or not the response came from cache. Unauthenticated probes and static assets are not counted; neither is traffic to this website.

Median response time

The time from a request reaching the billing gate to the response body being handed back to the web server, recorded on every row and summarised as the 50th and 95th percentiles. It is server-side: it does not include the network between you and us, or TLS. A figure measured from a client somewhere else will be higher, and how much higher depends on where the client is.

Responses without a server error

The share of responses in the window with a status below 500. A 4xx is the client's error and is counted as served; a 5xx is ours and is counted against us. This is not an uptime figure. A minute in which the API answered nobody because it was down leaves no rows and is therefore invisible here, which is why the day table shows volume alongside errors: a day with far fewer requests than its neighbours deserves a question.

Price freshness

The most recent fetched_at across all quotes from a source. A source is read on a schedule, nightly for Cardmarket and TCGplayer, so a figure of a few hours is normal and a figure of a few days means a feed has stalled.

What this page does not tell you

It has no external monitor and no incident history yet; the service has been logging since September 2026 and the window is capped at what exists. If you need a contractual number, the Scale plan carries a 99.95% uptime SLA.