DataShield Auth v1.1.8

Authorization and evidence for AI agents

Break-glass for agents: emergency access that expires itself, revokes mid-session, and can't be quietly deleted from the log.

Auth is the authorization and evidence layer for agent infrastructure: a Cedar policy engine with policy_explain for admin and platform decisions, scope-ceiling MCP tool tokens, per-tool-call metering attributed to agent identity, and a hash-chained audit log with signed checkpoints.

Break-glass lifecycle and scope ceiling Nested scope ceilings make escalation impossible by construction; a break-glass session is dual-logged, revocation-checked per call, and its evidence chain detects any deleted entry. SCOPE CEILING Owner scope Delegated key Session authority.ts ceiling escalation impossible by construction BREAK-GLASS TIMELINE break_glass_initiate dual-logged active session ◇ revocation? — every call revoked mid-session context downgraded next call TTL auto-expiry EVIDENCE LANE attempted delete Ed25519 checkpoint verify_chain → DELETION DETECTED log can’t be quietly deleted

How every governed tool call is authorized

Every governed tool call runs the same dispatch pipeline: token scope ceiling → declared authority tiermid-session revocation re-check → metered dispatch → sealed audit entry. The Cedar policy engine handles admin, platform, config, and token decisions — with policy_explain so you can see exactly why a decision went the way it did.

The pipeline is not a marketing diagram. The architecture page documents it end to end, stage by stage, as it exists in code.

See it live in the Demo Center →

Break-glass access for AI agents, delegation, and mid-session token revocation

Break-glass that expires itself

Emergency access is initiated as a dual-logged event, runs against a TTL, and auto-revokes. The initiation and every action taken under it land in the hash chain — an attempt to delete the entry is detected by verify_chain as a deletion, not silently absorbed.

Delegation with scope ceilings

A delegated key can never confer more than its owner intends — even for global admins. The ceiling is enforced by construction in the dispatch pipeline, not by convention.

Mid-session revocation

Revoke or suspend an agent and the change is caught on its very next tool call — the session context is downgraded mid-flight. Asana's 34-day cross-tenant MCP exposure would have been one call.

NHI authorization with metering, evidence, and governed configuration

Per-tool-call metering with agent attribution

Every governed call is metered and attributed to the agent identity that made it — usage accounting and authorization evidence come from the same pipeline stage.

Hash-chained audit with signed checkpoints

SHA-256 row chain with Ed25519-signed checkpoints, themselves chained. Verification distinguishes tampering, insertion, deletion, and truncation. GDPR crypto-shred erasure and ISO 27560 consent receipts are built in — erasure doesn't break the chain. Run the verification yourself.

Governed configuration and tenancy

Config changes go through propose/approve governance with security floors, encrypted connection credentials, and multi-tenant isolation across organizations, apps, and instances.

Three questions engineers ask us

Does Cedar authorize every tool call?

No — and we won't claim it does. Per-call authorization is the dispatch pipeline: token scope ceiling, authority tier, mid-session revocation re-check, then metered dispatch. Cedar fires on admin, platform, config, and token decisions, with policy_explain for auditability.

We already have Entra / Okta. Why Auth?

Bring your IdP. Entra or Okta tells you who the agent is — DataShield proves what it did was allowed. Auth is the authorization and evidence layer downstream of identity, not a replacement for it.

Can an admin quietly delete a break-glass entry?

Deletion is detectable. Audit rows are SHA-256 chained and sealed under Ed25519-signed checkpoints that are themselves chained — verify_chain reports deletion, insertion, tampering, and truncation as distinct verdicts.

Auth v1.1.8 — in production. Downstream of Auth: Ontology, the governed data plane; kept alive by Guardian.

Scope-ceiling authorization and tamper-evident evidence, self-hosted on your infrastructure.

See pricing

You've seen the proof

Ready for a number? Scope your deployment and we'll price it against your own economics.

Get your quote →