NEW  AI investigations now open fix PRs automatically — see what's new →
Built in the open · updated weekly

What we shipped.

No "Q3 roadmap" pages. No "Coming soon." Just the actual list of what we shipped, when, and what it does. Subscribe by following the GitHub repo.

Platform · Marketing

Marketing relaunch — agents, integrations, comparison, open-source pages

  • New /agents/ page introducing the 5 named agents (On-call SRE, Investigation, Cost FinOps, SLO Reliability, Deploy-aware Code).
  • Expanded /product/ page from 8 to 20+ documented capabilities.
  • New /integrations/ page with categorized native + MCP adapter grid.
  • New /compare/ page — 27-capability honest matrix vs Resolve.ai, Sherlock.ai, Metoro.io, Komodor, Cleric, Causely, k8sgpt.
  • New /open-source/ page with one-command Helm install and BYO-LLM story.
Demo · UX

Demo flow — in-page form, no calendar widget

  • New "Talk to us" modal — 4 fields, posts to backend SQLite, no third-party trackers.
  • Per-IP rate limit (5/hour) + hidden honeypot field for spam protection.
  • Authed admin page at /oneinfra/demo-requests to list, reply, mark replied.
Design System

Design system v2 — semantic color palette

  • Indigo demoted to brand-mark + signature gradient only.
  • Emerald owns security / your-VPC, amber owns cost / waste, rose owns incidents.
  • Primary CTA now near-black (Vercel-style max contrast).
  • Pulsing emerald "Runs in your VPC" dot (Stripe-style live indicator).
  • Per-stat colors: 73% amber, 100% emerald, 5× indigo gradient.
Backend · AI

V1 backend — 5 deferred picks

  • A1 — Real-time topology WebSocket stream with change attribution.
  • A6 — Eval Harness with 10 synthetic K8s cases + scoring + run history.
  • A7 — Recommended MCPs catalog (Honeycomb, Lightstep, Vector, Mezmo, Sumo Logic).
  • B1 — Trust scoring (heuristic V1; XGBoost slot at 1000+ rows).
  • B5 — k8sgpt HTTP analyzer plugin (point k8sgpt at oneinfra).
Cost · FinOps

Multi-cloud cost — AWS + GCP + Azure

  • Cluster cloud-detector (URL pattern → AWS / Azure / GCP / k3s).
  • GCP monitoring + billing adapter (google-auth-library + REST).
  • Azure cost adapter (SP OAuth2 + Cost Management API).
  • Cluster cards now show cloud badge (☁ AWS / Azure / GCP / k3s).
UI · Polish

UI polish — disabled clusters, eval harness sparklines, empty states

  • U3 — Disabled clusters render dimmed with Enable CTA.
  • U4 — Skeleton loaders + branded EmptyState everywhere first-load matters.
  • U5 — Eval Harness "Top failing cases" card + recent-runs trend sparkline.
Investigations · AI

Investigation Workbench + post-mortem auto-generator

  • Three-column workbench: AI analysis · pinned evidence · markdown scratchpad.
  • Generate post-mortem composes everything into a ready-to-paste Confluence doc.
  • Investigation thread tree shows parent / child investigations for follow-ups.
Graph · Knowledge

BlastRadius DAG + knowledge graph

  • Cytoscape-rendered DAG shows what else is at risk when an investigation lands.
  • Knowledge graph persisted across investigations — every fix teaches the next.
On-call · Incidents

On-call hardening — agent personas + Slack threaded replies

  • Verifier reranker keeps the model honest on low-confidence answers.
  • Slack investigation threads now post structured replies as the work progresses.
  • AI alert suppression + severity-aware grouping.
  • OTLP audit log — every action emitted as a standard OTel span.
Infrastructure · Multi-cluster

Federated multi-cluster + agent install

  • Add-cluster Flow A (kubeconfig paste) and Flow B (helm-installed agent for firewalled clusters).
  • Federated cross-cluster workload dependency detection.
  • Per-cluster cost fan-out + Holmes routing.
Want to see what's next?

Talk to us before we build it.

Design-partner beta is open. Your problem might already be on our list — or it might bump up the priority.

See the live app →
Talk to us

A 15-min conversation. No calendar tango.

Drop your details, we reply within 4 business hours. If a live walkthrough makes sense, we'll set one up — otherwise we'll just answer your questions over email.