- CollectionAgent Runtime & Infrastructure
- Classificationagent-native
- MCPUnavailable
- Agent SkillUnavailable
- InterfacesCLI · API
What it does
APort Agent Guardrails verifies AI agent actions before tool execution. It checks each proposed tool call or sensitive action against an Open Agent Passport policy and records allow/deny decisions, reducing reliance on prompt-only safety instructions.
Agent-facing interfaces
- Catalog protocols: CLI, API
- MCP: ❌ Not a standalone MCP server
- Agent Skill: ⚠️ No standalone Agent Skill listed in this catalog yet.
- Onboarding type:
cli
Protocol surface
| Interface | Detail |
|---|---|
| CLI | npx @aporthq/aport-agent-guardrails |
| Node runtime | npm package and framework-specific setup |
| Python runtime | PyPI package and framework-specific callbacks/providers |
| API mode | Optional hosted policy evaluation through APort API |
| Local mode | Offline policy subset for local/bash guardrail paths |