Security at the agent boundary
A safer doorway for agent actions.
usezn is an AI security gateway for agents that call tools through MCP, A2A, and ordinary HTTP APIs. We exist because an agent can be technically correct and still be unsafe: untrusted text can arrive through a document, web page, email, or another agent and attempt to change what the agent is allowed to do.
Our gateway evaluates the tool-call input before execution. It returns an allow or block decision, the matched rule, confidence, and an evidence identifier. The goal is not to replace identity, authorization, or human review. The goal is to add a fast, explicit policy boundary between an agent and the systems it can affect.
usezn is built for teams that want a small integration surface and an audit trail they can understand. The managed API provides a deterministic hot path, a local playground, MCP integration, and evidence records with documented retention. The open documentation describes the limitations as clearly as the capabilities.
Protect the boundary
Inspect untrusted instructions before a tool call reaches your systems.
Make decisions legible
Return a verdict, reason, rule, confidence, and evidence ID instead of a silent guess.
Keep integration small
Use REST, MCP, the CLI, or the local playground without rebuilding your agent loop.
The people and the record
usezn is maintained by a small independent team. Product decisions, pricing, API behavior, privacy commitments, and integration limits are documented publicly so developers and agents can inspect the same source of truth. We welcome reproducible security reports and actionable feedback.