Repository sovereignty
Analysis and verification start in your environment. Cloud upload is opt-in. High-Assurance fails closed on non-local paths.
RIVA is the AI-native control plane for trusted delivery: local-first by default, gated agent change, and standards-aligned evidence — including a High-Assurance profile for security-sensitive teams.
Govern · Prove · Assure · Ship
Add retry with jitter on the checkout webhook handler. Scope limited to src/webhooks/ and its tests. Two callers stay outside the blast radius.
High assurance starts with control. Source stays local. Agents stay in scope. Advance only with proof.
Read the Trust page →Coding agents write. RIVA decides what is safe to advance.
One sequence from characterized HEAD to a pack your risk team can read.
Characterize HEAD. Pin what “green” means before agents accelerate.
Blast-radius map. Agents stay inside the boundary.
Unit · Integration · Journey. A person approves.
Standards-tagged pack for review and CI.
Analysis and verification start in your environment. Cloud upload is opt-in. High-Assurance fails closed on non-local paths.
Prepare → Agent → Gate → Advance. The model’s reply is never the definition of done.
Export what changed, what stayed green, who approved — mapped to control themes security teams already use.
Enforce local-first defaults, bound agent scope, require human gates, and export evidence of what changed and what stayed green.
Local-first analysis and verification. Bring your own private or on-prem models.
Blast-radius scope, gated loops, and a shared Unit · Integration · Journey bar.
Provenance-backed verification and human authority at lock and release.
Product control alignment — not organizational certification.
One contract across the editor you already use, MCP, and CI.