--- title: Infrastructure description: The infrastructure behind your alerts: a multi-provider probe fleet across the globe, an EU control plane in a TIER III data center, and cryptographically verifiable probes. canonical: https://watchfor.io/infrastructure --- # The infrastructure behind your alerts A monitoring platform has one job above all others: be more reliable than the things it watches. This page shows how WatchFor is built — the probe fleet, the control plane, and the engineering decisions that keep a false alarm from ever paging you. 20 probe locations 18 countries 7 world regions 3 independent providers TIER III control-plane facility 100% infrastructure as code ## A probe fleet that doesn't share a fate 20 probe locations across 18 countries run your checks from where your users actually are — every one dual-stack (IPv4 + IPv6), every one listed with its exact addresses at [/ips.json](/ips.json). Three independent providers. The fleet spans AWS regions on six continents plus two separate Lithuanian network operators (Telia and Interneto vizija) — so even a whole-provider incident can't blind the region, and geo-restricted sites are covered from truly local paths. Provably ours. Every probe has forward-confirmed reverse DNS under *.probe.watchfor.io and signs every HTTP request with the fleet's Ed25519 key (RFC 9421 Web Bot Auth) — verify any request against our [published keys](/.well-known/http-message-signatures-directory). Details on [the bot page](/bot). ## An EU control plane on serious ground Scheduling, alerting, storage and the dashboard run on dedicated enterprise hardware in a TIER III–certified Telia data center in Vilnius, Lithuania — the largest open data center in the region. Your monitoring configuration and incident history live in the EU, under European law, with GDPR applying end to end. ### TIER III facility Concurrently maintainable: N+N UPS power, N+1 diesel generators, N+1 climate control, 24/7 on-site security and CCTV. ### EU data, EU law Monitoring configuration and incident history stay in the European Union, under European law — GDPR applies end to end, with carrier-grade DDoS protection upstream. ### Dedicated hardware Enterprise Dell servers we operate ourselves — no resellers, no noisy neighbours on the control plane. ### Everything as code Every server — probes included — is provisioned and configured from version-controlled Ansible. No hand-built snowflakes to drift or to lose. ## Certified ground, green power 100% renewable energy The data center runs entirely on renewable electricity — every check we run for you is powered by green energy, certified under ISO 14001 (environment) and ISO 50001 (energy management). Facility certifications TIER III ISO 27001 — information security ISO 27017 — cloud security ISO 9001 — quality ISO 20000-1 — IT service management ISO 14001 — environment ISO 50001 — energy ISO 45001 — occupational safety Independently audited certifications held by the Telia facility that hosts our control plane — the same bar banks and telecoms require of their infrastructure. ## How an alert is born - ### 1 · Checks run at the edge A Go probe executes your check — HTTP, DNS, SSL, ping, MTR, an MCP session — from the locations you chose, on the interval you chose. - ### 2 · Results stream home Every result travels over a TLS message bus (NATS JetStream) with at-least-once delivery, into a columnar time-series store (ClickHouse) built for billions of checks. - ### 3 · Failures must agree One failed check marks a monitor Degraded — nothing pages. Only when several independent locations confirm the failure does it become a Down incident. - ### 4 · The alert carries the cause The incident ships with the exact rule that fired, per-location evidence and timing breakdowns — to email, Slack, PagerDuty and 13+ more channels. ## 100% agent-ready — AI can operate WatchFor A growing share of monitoring is set up by AI assistants acting for their users — and WatchFor is built for that. An agent can discover the platform, authenticate with scoped OAuth tokens and manage monitors end to end with no human in the loop. Independent agent-readiness scanners grade exactly this, and we treat their audits like uptime: publicly, continuously. Independently scored The full agent stack, in production MCP server — 45 tools Docs MCP server A2A agent — 16 skills OAuth 2.1 + PKCE REST API + OpenAPI 3.1 llms.txt + agents.md No-auth sandbox Signed probes (RFC 9421) SDKs: npm · PyPI · RubyGems Add watchfor.io/api/mcp to Claude or any MCP client and it gets scoped access through an OAuth consent screen — no API key ever copied. Everything an agent can do runs on the same authenticated API, rate limits and audit trail as the dashboard. Start at [/agents.md](/agents.md). ## Reliability engineering, not reliability marketing - ### Probes hold no secrets Each probe authenticates with its own least-privilege credential that can publish only its own results — a compromised edge node can't read anyone's data or impersonate another location. - ### Isolated by design Probes are deliberately cut off from the control plane's internals: they receive check commands and return results over one authenticated, TLS-encrypted channel — nothing else. - ### The scheduler holds no fragile state Check schedules are computed, not stored — the engine was load-tested to 25 million active checks on a single node, and we've written up how it works on the engineering blog. - ### We eat our own dog food watchfor.io is monitored by WatchFor, from this same fleet, with the same confirmed alerting — when something degrades, our own pagers go off first. Want the deep technical version? Read [how the scheduler handles 25 million checks](/blog/scheduling-25-million-checks-one-go-binary) and [how we made the platform agent-ready](/blog/what-it-took-to-make-monitoring-agent-ready). Security disclosures: [security.txt](/.well-known/security.txt). ## Put this fleet to work on your infrastructure Free plan, no credit card — your first monitor is checking from around the world in about a minute. [Start free](/auth/sign-up)[Verify our probes →](/bot) --- Canonical page: https://watchfor.io/infrastructure · Site guide: https://watchfor.io/llms.txt