Reliability

Built to stay up - and to recover fast

Keel is where your compliance program lives, so availability and recoverability matter as much as features. Here's how the platform is engineered today - described plainly, with no promises we can't back.

Always-on delivery

The application runs on Cloudflare's global edge network - served from 300+ cities and routed to a healthy location automatically. The app layer is stateless, so there's no single server whose failure takes it offline.

Durable, recoverable data

Your data lives in managed Postgres (Neon) with point-in-time recovery, and evidence files sit in object storage (Cloudflare R2) engineered for 99.999999999% durability. Everything is encrypted in transit and at rest.

Fast, safe recovery

Every release is versioned and can be rolled back instantly. Errors are monitored continuously and core dependencies are health-checked, so problems are caught and reversed quickly.

Availability & redundancy

Global edge, automatic failover. The app and marketing site run on Cloudflare Workers and Pages across a worldwide network. Requests use anycast routing, so if a location is degraded, traffic is served from another automatically - there's no single data center you depend on to reach the app.
Stateless, horizontally distributed app. Application servers hold no durable state, so capacity scales out and any instance can serve any request. A failed instance is simply routed around.
Versioned deploys with instant rollback. Releases are immutable versions. If a deploy misbehaves, we roll back to the previous known-good version in minutes.

Data durability & backup

Point-in-time recovery. The database (Neon Postgres) supports restoring to an earlier moment, so an accidental change can be undone rather than lived with.
Off-provider encrypted backups. Nightly, the database is dumped, encrypted, and copied to separate object storage - a second, independent copy that survives a provider-side failure - and an automated monthly drill restores it to prove the backups are actually recoverable.
High-durability evidence storage. Uploaded evidence is stored on Cloudflare R2, which is engineered for eleven 9s (99.999999999%) of annual durability.
Encryption end-to-end. All traffic is served over HTTPS/TLS, and data at rest is encrypted by our infrastructure providers.
Tenant isolation. Every workspace is isolated in the database with Postgres row-level security - enforced in the database, not just the app. See our security overview for details.

Monitoring & recovery

Continuous error monitoring. Application errors are captured and traced so issues surface fast and are resolved quickly.
Automated health checks. Core dependencies are probed by a dedicated health endpoint, suitable for external uptime monitoring.
Documented DR/BCP runbooks. We maintain internal disaster-recovery and business-continuity runbooks with defined recovery objectives (RPO/RTO) and step-by-step restore procedures - and we rehearse recovery rather than assuming it works.

We run our program on the same platform

Keel isn't just the tool - it's how we run our own compliance and resilience program. Our controls, evidence, and vendor reviews live in Keel, mapped to the frameworks we support. As our program matures we'll publish more here.

Want the current operational picture? Check our system status, read the security overview, or email [email protected] with questions.

Start free Security overview →

Cloudflare, Neon, and other provider names are referenced factually and are the property of their respective owners. Durability figures reflect the stated engineering targets of the underlying providers.