Control Plane

Aegis

"Is this action safe?"

Intercept, validate, and enforce every agent action before it executes — even if the agent has been compromised.

Design principle: assume breach.

Live Decisions

aegis.decisions.live

[ALLOW] salesforce.update_contact → field: phone

[DENY] salesforce.bulk_delete

reason: count=1,432 exceeds cap=100

[REWRITE] hubspot.create_deal

amount: $2.4M → $500K (ceiling enforced)

[DENY] db.raw_query — pii_columns_blocked

✓ System safe. 3 dangerous actions blocked.

✓

Policy Decision Point

ALLOW / DENY / REWRITE / WARN on every agent action. Fail-closed — if policy cannot evaluate, the action is blocked.

✓

JSON Predicate Engine

10 operators, compound all/any rules, glob matching, 60s cache. Evaluates in microseconds.

✓

Formal SHACL Constraints

Business rules as formally verifiable graph shapes — not just best-effort policy.

✓

Action Rewrite Engine

Cap values, mask fields, set defaults before execution — without blocking the action.

✓

Signed Attestation Store

HMAC-SHA256 signed at creation. Append-only — no record can be modified after the fact.

✓

Identity Adapter

ABAC trust levels. AWS IAM, Azure MSI, and SPIFFE/SVID workload identity integration.

✓

FinOps Token Gating

Hard cost ceilings per agent, per workspace, per month. Runaway LLM bills are impossible.

✓

SDK Wrappers

Drop-in for LangGraph, AutoGen, and CrewAI. One line of code to protect your agent swarm.