Products
Our Product System
Access, actions, and validation are complementary control points, not interchangeable ways to describe the same product.
Axora Security organizes enterprise AI security around access, actions, and validation. These are complementary control points, not interchangeable ways to describe the same product.
Compare The Products#
| Axora Router | Axora AgentGuard | Axora Arena | |
|---|---|---|---|
| Primary question | Who can use AI, through which model channels, and under which data policies? | Can this component be trusted, and should this agent action proceed? | What can an adversary do within the authorized test scope? |
| Main control point | AI API access and request/response flow | Agent components and supported runtime execution paths | Scoped adversarial testing and security assessment |
| Main objects | Users, groups, applications, model channels, quotas, and content | Agents, skills, plugins, MCP tools, knowledge-base connections, credentials, and permissions | Authorized target systems and connected workflows |
| Typical evidence | Request decisions, information-flow traces, and usage attribution | Asset inventories, scan reports, policy decisions, approvals, and event timelines | Findings with stable IDs, attack paths, and reproducible evidence |
| Primary teams | Enterprise IT, AI platform, security, and compliance | AppSec, platform engineering, DevOps, data security, and audit | Security, product security, red team, and penetration testing |
How The Control Points Complement One Another?#
Consider an enterprise coding agent that uses an external model and tools on a development host:
- Access. Router provides a managed path for model API usage. Group policies determine how requests and responses are handled, while identities and quotas organize usage.
- Actions. AgentGuard reviews installed components and evaluates supported tool calls on the agent host. A risky action can generate an alert, require human approval, or be blocked according to policy and integration capability.
- Validation. Arena tests an explicitly authorized target and records findings and evidence. Teams use that evidence to prioritize fixes and retest the relevant attack paths.
Where The Boundaries Differ?#
Gateway policy is not host-level action control#
Router evaluates information flowing through its gateway. An agent's local command execution, file access, or tool use is a separate control point addressed by AgentGuard on supported integrations.
Component review is not runtime approval#
A reviewed skill or plugin can still request a risky action. AgentGuard combines pre-adoption review with action evaluation so component trust does not replace runtime policy.
Testing is not continuous enforcement#
Arena produces evidence from authorized assessment. Router and AgentGuard apply policies at their respective control points. A test finding informs security decisions; it does not itself enforce a production policy.
Visibility is not blocking#
Discovering an agent does not establish that its actions can be intercepted. AgentGuard distinguishes full, partial, observation-only, and unsupported coverage according to host capabilities.
Choose a starting point#
| Your immediate need | Start with |
|---|---|
| Organize employee or application AI access and sensitive-data handling | Axora Router |
| Review third-party agent components before adoption | Discovery and Review |
| Evaluate commands, tool calls, credential access, or agent data egress | Runtime Protection |
| Establish AI asset visibility, approvals, and audit evidence | Governance, Threat Intelligence and Validation |
| Validate an authorized target and reproduce security findings | Axora Arena |