v2.6May 06, 2026
platformpolicy
Policies as code, now first-class.
You can now author tenant policies in a typed YAML dialect, attach them to detectors, and version-control them in your own repo. The platform compiles them at deploy and surfaces conflicts before the policy goes live.
- New Policy tab under each workspace with diff view against last deployed version.
- Git sync (GitHub, GitLab self-hosted) under Settings → Integrations.
- Conflict surfacing: overlapping policies are flagged with the offending rules quoted side-by-side.
v2.5Apr 22, 2026
detectors
Live map shows tool, agent, decision.
The monitoring view now stitches together each AI action across the three layers — which tool ran, which agent orchestrated it, which decision it produced. Hover any node to see the upstream chain.
v2.4Apr 04, 2026
explainability
Decision receipts.
Every decision the platform surfaces now ships with a receipt: the inputs the model saw, the policy it satisfied, the human who reviewed it (or didn't). Exportable as signed JSON for audit.
v2.3Mar 18, 2026
integrations
Slack & Linear, native.
Native integrations for Slack (incident routing, daily digests) and Linear (auto-file follow-ups from incidents). OAuth setup under Settings → Integrations; no zaps or middleware.
v2.2Mar 03, 2026
platformperformance
Faster monitoring backbone.
Rewrote the event-ingest pipeline on a streaming backbone. P95 monitoring latency went from 4.2s → 480ms across tenant sizes. No action needed on your side; existing detectors run unchanged.
v2.1Feb 14, 2026
security
SCIM, SSO, and tenant-isolated keys.
SCIM provisioning for Okta and Azure AD. SAML SSO is now standard on every plan, not enterprise-only. Per-tenant encryption keys (BYOK) available on request — keys held in your KMS, never ours.
v2.0Jan 30, 2026
major release
Ovna 2.0 — the trust layer.
The platform rewrite we've been talking about since the autumn. Three pillars now generally available:
- Visible. A live map of every AI action across the stack.
- Answerable. Decision receipts and explainability built-in, not bolted on.
- Safe. Policy-as-code, encoded once, enforced everywhere.
Migration from 1.x is handled automatically; existing detectors are translated on first deploy. Open a ticket if anything looks off.