Reads your repos. Measures your vendors independently. Alerts only when something of yours breaks, then computes what you're owed.
Free to start. Zero config.
● Live from the incident engine
All incidents →How it works
696 package-to-service mappings built in. We scan lockfiles, manifests, Dockerfiles, Terraform, and CI pipelines across 25 ecosystems. No manual config.
Add our free SDK for real-time production telemetry. TypeScript, Python, Go, Ruby, PHP, Rust, Java, .NET. Five lines of code. 1,000 events/min on every plan.
Alerts cover the services your projects actually depend on — vendor incidents outside your stack stay out of your pager. Independent signal streams are weighted against each other, so no single noisy channel can over-fire; when they corroborate, we promote the incident — before the vendor acknowledges it.
When vendors breach their published SLA, we calculate the credit tier and generate an evidence package you can file with one click. Alerts go out through 12 channels the moment something goes wrong.
SLA termsAuth0 < 99.99% uptime5% creditOne real detection · 2025-11-04
Cloudflare Workers AI browned out for teams running bge-m3 embeddings. This is what the engine saw, timestamp by timestamp — and when the official status page finally said anything.
Every timestamp above is from the engine's own records of that incident — the kind of evidence a credit claim gets built on.
Read the full teardown →T+0:00
Three community posts on Bluesky and Hacker News inside 90 seconds: "bge-m3 inference calls returning 502 on Workers."
T+0:01
Cloudflare Radar signal goes yellow for Workers AI in North America.
T+0:03
Composite score crosses the promotion threshold. Incident promoted — 17 customer projects flagged degraded, alerts dispatched.
T+0:07
Cloudflare’s status page posts: "Investigating elevated errors on Workers AI."
Watching the services you build on
232 services tracked around the clock — any logo opens its live reliability page.
All product names, logos, and brands are property of their respective owners. Use of these names, logos, and brands does not imply endorsement.
Platform
Blast radius
See which projects, features, and teams are affected. Cascade analysis follows dependencies 3 levels deep.
What-if simulation
Model hypothetical outages. Get a ranked list of affected projects with confidence scores per region.
Risk scores
0–100 per service. Weighted across uptime, error rates, incident frequency, and resolution time.
MCP for AI tools
11 tools for Claude, Cursor, and other agents. Check blast radius, diagnose errors, audit dependencies.
CLI + API
Health, risk scores, and incidents from your terminal. Full REST API for automation and scripting.
Vulnerability alerts
Cross-references OSV and GitHub Advisory databases. Correlates CVEs with service health data.
25
Ecosystems
232
Services
12
Alert channels
FAQ
No. Alerts cover the services your projects actually depend on — the dependency graph mapped from your repos. A vendor incident that doesn't touch your stack doesn't page you; it stays visible on the public live pages if you go looking, but silence is the default.
We watch independent signal streams for every service: the official status page (polled on a 1–5 minute cadence), vendor RSS feeds, synthetic probes hitting real endpoints, and Cloudflare Radar. Optional SDK telemetry adds your own production error rates as a further signal. When independent streams corroborate, we promote an incident without waiting for the vendor to acknowledge it.
Two projects, five monitored services, every alert channel, and SDK telemetry at 1,000 events per minute. No credit card. Paid plans raise the limits and tighten the polling cadence.
We track the uptime commitments vendors publish. When measured downtime crosses a credit tier, we calculate what you're owed and assemble an evidence package — incident timeline, status-page snapshots, and your SDK error rates — ready to file inside the vendor's filing window.
No. Connect a repository read-only and we map your lockfiles, manifests, Dockerfiles, Terraform, and CI pipelines to the services they depend on. The SDK is optional — five lines of code if you want production telemetry.
The stack you don't run yourself: cloud platforms, payment APIs, AI providers, databases, and the tools in between — 232 services across 25 package ecosystems, each with a public reliability page on the leaderboard.
Free for 2 projects and 5 services. No credit card. Connect a repo, see your service health in under five minutes.