Internet infrastructure monitoring for engineers

See your infrastructure the way the internet does.

ProbeAtlas measures availability and DNS, TCP, TLS, and HTTP timings of your public endpoints from monitoring agents on independent cloud networks, and tells you whether a slowdown is global, regional, or one network's problem.

Sign-up is by invitation while we work with our first design partners.

Five questions, answered with real measurements

Is it reachable everywhere?

Availability of each endpoint from every monitoring location, with failures categorized instead of lumped together.

Where is it slow?

Latency per region side by side: minimum, P50, P95, and P99, computed only when there are enough samples to mean something.

Which layer is degrading?

DNS resolution, TCP connect, TLS handshake, time to first byte, and total time are measured separately on every HTTP check.

Global, regional, or one network?

Incidents are classified by where they happen, so one provider's routing problem does not look like an outage.

What changed over time?

Raw results are kept with both the agent's measurement time and the server's ingestion time; gaps are shown as missing, never as success.

Where we measure from

This list comes live from the platform: the agents it runs right now, with the location their operator configured. Coverage grows as locations are added; we never infer locations from IP addresses.

No monitoring agents are enrolled yet.

Checks

HTTP and HTTPS

Status code against your expected codes, DNS, connect, TLS, time to first byte, total time; redirects followed and re-validated.

DNS

Resolution time and answers for A, AAAA, CNAME, MX, TXT, and NS records, optionally against a specific nameserver.

TCP

Connection time to any public port.

ICMP

Round-trip time and packet loss, where the agent's host allows unprivileged ping.

API-first, from day one

Everything the dashboard does goes through a versioned REST API with an OpenAPI spec, API keys you can rotate, and a CLI with JSON output, so monitors live next to the rest of your configuration.

curl -s https://api.<your-deployment>/v1/monitors \
  -H "Authorization: Bearer $PROBEATLAS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name": "api", "type": "http", "target": "https://api.example.com/health",
       "regions": ["eu-west", "us-east"], "interval_seconds": 60}'

probeatlas monitors list            # the CLI prints JSON too
  • Fields that were not measured are null, never zero.
  • Percentiles need enough samples (P95 at least 20, P99 at least 100); otherwise they are left out.
  • Every workspace is isolated; API keys are stored hashed and shown once.
  • Agents only contact public addresses and re-check every redirect.