Skyrelis is an AI security platform that provides always-on runtime security for LLM-driven, multi-agent workflows through behavioral monitoring, centralized policy enforcement, and protocol-layer inspection. Unlike developer-focused AI security tools, Skyrelis is built specifically for security operations teams who need centralized oversight of agent deployments.
The company was founded by Jaz Lin, a network security engineer who spent seven years at Cisco earning her CCIE certification before holding product and security roles at VMware, Rubrik, and RingCentral. Lin founded Skyrelis after identifying a gap in the market: existing AI security tools were built for developers, not for security operations teams who need centralized oversight of agent deployments.
Skyrelis is currently pre-Series A, working with design partners and early production deployments. The team is deliberately small and focused on building for enterprise security operations use cases.
What is Skyrelis?
Skyrelis monitors and governs AI agents at runtime through a combination of a lightweight SDK and a centralized security gateway. Every agent action passes through the gateway for policy enforcement, risk scoring, and audit logging.
The core insight behind Skyrelis is that security policies should be managed centrally by security teams, not embedded into individual applications by developers. The platform provides a centralized portal where security operators can see all agents across the organization and configure policies from a single interface.
Key Features
| Feature | Details |
|---|---|
| Deployment Model | Lightweight SDK + centralized security gateway |
| Monitoring | Always-on behavioral monitoring with contextual risk scoring |
| Policy Enforcement | Centralized portal for managing policies across all agents |
| Protocol Support | MCP (Model Context Protocol) and A2A (Agent-to-Agent Protocol) compatible |
| Prompt Security | Intercepts and blocks prompt injection attacks at runtime |
| Data Protection | Blocks sensitive data exfiltration attempts through agent channels |
| Audit Trails | Full logging of every agent interaction with risk context |
| Target Users | Security operations teams (not developer-focused) |
| Stage | Pre-Series A, design partner phase |
How the security gateway works
Skyrelis routes all agent communication through a centralized security gateway. The gateway sits between agents and their target tools, data sources, and other agents, inspecting traffic at the protocol layer.
When an agent initiates an action, the gateway:
- Intercepts the request — Captures the action before it reaches the target system
- Evaluates against policies — Checks the action against centralized security policies configured by the security team
- Applies risk scoring — Assigns a contextual risk score based on the agent’s identity, the action type, and the target resource
- Enforces controls — Blocks, allows, or flags the action based on policy rules and risk thresholds
- Logs the interaction — Records the full interaction with risk context for audit trails
This approach means security teams do not need to modify individual agent applications to enforce security policies. The gateway handles enforcement transparently at the protocol layer.
MCP and A2A compatibility
Skyrelis is built for interoperability with emerging agent communication standards. The platform supports the Model Context Protocol (MCP), which provides a standardized interface for AI applications to connect with external data sources and tools. It also supports the Agent-to-Agent Protocol (A2A) for inter-agent communication.
This protocol-layer compatibility means Skyrelis can inspect and enforce policies on agent interactions regardless of the underlying agent framework, as long as the agents communicate through supported protocols.
Built for security operations
A deliberate design choice in Skyrelis is the target user. Where many AI security tools are developer-focused, requiring code changes or SDK integration by engineering teams, Skyrelis prioritizes the security operations workflow. The centralized portal lets security teams:
- View all discovered agents across the organization
- Configure and manage policies without code changes
- Monitor agent behavior and risk scores in real time
- Investigate incidents with full audit trail context
Getting Started
When to use Skyrelis
Ideal for organizations running multi-agent AI workflows that need centralized security oversight managed by security operations teams rather than developers. The protocol-layer approach works well in environments using MCP and A2A standards.
The platform is best suited for enterprises that want to separate security policy management from application development — where security teams define the rules and the gateway enforces them without requiring changes to individual agent applications.
For more AI security tools and guidance, see the AI security tools category page. For LLM vulnerability scanning, look at Garak or Promptfoo. For runtime prompt protection, consider Lakera Guard or LLM Guard. For enterprise AI governance, see Onyx Security or Noma Security. For agent access control, check Alter.
