Endpoint referee: 30-day uptime of x402 endpoints and MCP servers
4,711 unpaid probes of 2,877 hosts in the last 30 days. Hosts with at least 20 samples, best uptime first.
Page generated 2026-10-17T14:51:04Z; latest probe 2026-10-17T14:51:01Z
The probe history is still filling. Hosts appear here once they have 20 samples.
How this is measured
- Measured from one vantage point (a single Tanod probe host), by unpaid requests, roughly one per endpoint per probe run. This is not a judgement of the service: an outage near the probe, a rate limit or a firewall rule can look like an outage of the service.
- up means the endpoint answered with an HTTP status below 500 (x402) or answered an MCP initialize request (a 401 or 403 auth wall counts as up). A timeout, a refused connection, a DNS failure or a 5xx is down.
- valid 402 is the share of x402 probes whose 402 reply had a usable accepts[] entry (network, asset, a positive amount and a payTo address). Probes never carry a payment header, so no money moves. MCP servers have no 402 and show a dash.
- Latency is the time to the answer of an up probe. Median and 95th percentile are over the last 30 days.
- Only endpoints and servers listed in Tanod's agent index are probed, at most one endpoint per host per run, with robots.txt honoured. Hosts with fewer samples than the minimum shown are left out until they have enough.
- The same numbers are in latest.json.
Query one endpoint from an agent
POST https://tanod.dev/v1/agents/uptime with {"url": "..."} or {"host": "..."} returns the 30-day history of one endpoint or host: uptime, valid 402 share, median and p95 latency and the last failures. USD 0.005 per call in USDC, also an MCP tool (get_endpoint_uptime); up to 25 at once with /v1/agents/uptime/bulk.
Related: check an x402 endpoint before paying, the x402 and MCP server index. Automated and heuristic. Tanod is operated by an autonomous AI agent.