Secure execution layer
for AI agents in the enterprise.
Aether sits between your agents and the world — enforcing policies on every tool call, logging every decision, and escalating to humans when the stakes are high.
AI agents in production are ungoverned by default.
Every agent framework ships with capability. None ship with policy enforcement, audit trails, or human oversight. That gap is your compliance risk.
Agents act without authorization
Your AI agent reads a database it should not touch, sends an email without approval, or calls an external API unchecked. No one knew until the damage was done.
No audit trail for compliance
Your auditor asks what the agent did last Tuesday. ISO 42001 and SOC 2 require more than a log file.
Human oversight is manual chaos
High-risk steps get reviewed by whoever happens to be watching Slack. No SLA, no routing, no escalation path.
Everything your agents need. Nothing they should not have.
Six capabilities, one policy boundary.
Policy Engine
Every tool call, every data access, every agent action evaluated against your Open Policy Agent rules before execution. Deny, allow, or escalate in under 15ms.
MCP Tool Control
Agents connect to the outside world through Model Context Protocol servers. Aether sits between agent and tool, enforcing policy on every single invocation.
Immutable Audit Trail
Every workflow, every policy decision, every agent action logged to an append-only structured audit log. Compliance evidence generated continuously, automatically.
Human-in-the-Loop
High-risk steps pause for human review before proceeding. Escalation routes to your team via Slack, email, or webhook. The agent waits.
Workflow Orchestration
Durable, multi-step agent workflows powered by Temporal. Steps run in sequence or parallel. Failures retry automatically. State survives restarts.
Multi-Tenant Isolation
Per-tenant MCP tool namespacing, isolated policy sets, and separate audit streams. Run dozens of customers on one deployment without data bleed.
Five steps. Zero unauthorized actions.
From agent tool call to audit log entry — in under 15ms.
Agent emits a tool_call — read database, send email, call API.
Aether's proxy captures the request before it reaches the tool.
Policy engine checks agent identity, tenant, tool, and context.
Approved calls proceed. Denied calls error. Flagged calls pause for human review.
Every decision — allow or deny — written to the immutable audit log.
The frameworks your auditors already know.
Aether maps directly to AI governance frameworks.
AI management system standard. Audit trail, policy engine, and human-in-loop controls map directly.
Govern, Map, Measure, and Manage functions covered by Aether's architecture.
PHI never leaves the enclave. Agent access to clinical data requires explicit policy allow.
Immutable audit log, least-privilege access, automated incident escalation. Control evidence generated continuously.
Start secure. Scale with confidence.
All tiers include the full policy engine, audit trail, and Keycloak integration.
No security features paywalled.
Team
Starter
Up to 5 agent types · 10K workflow runs/mo
- Policy Engine
- Full Audit Trail
- Keycloak Integration
- Email Support
Platform
Professional
Unlimited agent types · 100K workflow runs/mo
- Policy Sync
- Full Audit Trail
- Keycloak + SSO/SAML
- Self-hosted Langfuse
- Human-in-Loop Controls
- Priority Support
Regulated
Enterprise
Unlimited · Dedicated policy cluster · 99.9% SLA
- ISO 42001 Control Mapping
- NIST AI RMF Alignment
- Air-gap / VPC Option
- Compliance Report Exports
- Dedicated Support Engineer
- SLA + BAA Available
On-Premise
Self-Hosted
Your data center · Your infrastructure · Custom SLA
- Full Source Deployment
- Air-Gapped Option
- Your Keycloak Instance
- White-Glove Setup
- 24/7 Dedicated Support
- Government / Defence Ready
Your agents should have a paper trail.
Onboarding design partners in financial services, healthcare, and regulated SaaS — production by end of April 2026.