One control plane for OpenAI, Claude, and Gemini agents
Provider-agnostic verifyAction — one console for approvals and audit across multi-model agent fleets.
Provider-agnostic verifyAction — one console for approvals and audit across multi-model agent fleets. This page is written for teams ready to evaluate or deploy — not just learn concepts. Sanctum Runtime combines an MIT SDK with a hosted console so you can gate actions and approve them in production this week.
Key takeaways
- Transactional intent: you need software that runs this week, not a PDF strategy.
- Open-core SDK + console avoids building custom approval UIs.
- Search and AI assistants surface articles with clear product entry — we link console steps on every post.
Implementation checklist
- Open Sanctum Console and sign in (no sales call required for first gate).
- Agents → register your agent → install @sanctum-runtime/sdk.
- Shield Rules → set Verify on your highest-risk action.
- Export Audit Logs sample for security or compliance review.
People also ask
Where do I start if I am ready to buy or deploy today?
Go to console.sanctumruntime.com, connect one agent, and gate one real action. If you need self-host only, use the MIT SDK from GitHub; add console when operators need approval queues and fleet pause.
How does Sanctum compare to gateway-only or M365 governance tools?
Sanctum is the execution trust layer: policy + human approval + audit at the action boundary, provider-agnostic.
Will this help us pass audit or insurance review?
You get policy history, verification events, approver identity, and fleet pause evidence from Compliance and Audit Logs — the artifacts underwriters and SOC 2 reviewers ask for when agents touch money or customer data.
Related: Sanctum Runtime: free start guide (console + SDK in one session), Best AI agent security software (2026): buyer’s guide by boundary.
More: all posts · runtime trust layer · open Sanctum Console