Frequently asked questions

Agentic SRE is autonomous site reliability engineering: an AI that investigates an incident, proposes a fix, checks it against your SLOs and policies, then remediates — the way a senior on-call engineer would, but in seconds and without the 3 a.m. page. Warble Cloud's Reflexion engine runs this loop continuously, and every action is policy-gated and reversible.

The Living Flock is Warble Cloud's fleet of cooperating agents that run inside your own cluster: Reflexion (self-correcting Actor/Critic remediation), Starling (the Kubernetes platform), and Warble Brain (model serving and GPU scaling). Together they observe, decide, and heal — birds of a feather working as one operations team.

Reflexion observes the incident (metrics, logs, cluster state), the Actor proposes a remediation hypothesis, the Critic validates it against SLO impact and policy, and only approved actions execute via GitOps or kubectl. It self-corrects across multiple cycles before acting, and every outcome is recorded as a training signal — so the system improves with each incident.

Yes. Warble Cloud runs entirely in your own VPC or cluster, GitOps-native, with zero vendor lock-in. Your data, runbooks, and incident history never leave your perimeter — you own the stack and can uninstall at any time.

Starling exposes a Model Context Protocol (MCP) surface so AI assistants and tools can safely query and act on your platform — discovering services, reading metrics, and triggering governed actions through a single, policy-controlled interface.

Seat-based pricing at $300 per seat. Most teams are production-ready on their own cluster in about five working days, with no credit card to start and cancel-anytime terms.

Every proposed action passes a Critic that validates it against your SLOs and OPA-style policies, with a pre-computed rollback stored before anything executes. High-risk actions can require human approval, and the whole decision — context, reasoning, verdict, result — is recorded for audit and replay.

Yes. Reflexion reads from the tools you already run — Prometheus, Loki, Alertmanager and Kubernetes itself — and executes through GitOps or kubectl. There's no rip-and-replace; it sits on top of your current stack.

Avirka is a self-hosted SRE LLM family fine-tuned on operational incidents, served in-cluster (Ollama / vLLM) with an optional managed fallback. Because inference runs inside your perimeter, the platform can operate in air-gapped and sovereign environments.