AI Agent Guardrails for Every Framework
Pre-tool-call authorization that works with your existing stack. One command setup. Policy-based tool authorization with OAP passports. Open source.
Agent Frameworks
IDE & Code Editors
Workflow Automation
How Guardrails Work
Create a Passport
Define your agent's identity, capabilities, and operational limits in an OAP-compliant passport.
Install & Configure
One command installs the guardrail and writes your config. Works with pip, uv, or npx.
Every Tool Call is Authorized
Before any tool executes, the guardrail evaluates it against your passport's policy. Denied calls return structured errors.
Frequently Asked Questions
What frameworks does APort support for AI agent guardrails?
APort supports LangChain, CrewAI, DeerFlow, Cursor, Claude Code, n8n, and OpenClaw. Any framework with a tool execution hook or middleware system can integrate with APort's generic OAP provider.
How is APort different from Guardrails AI or NeMo Guardrails?
APort provides pre-action authorization (before tools execute), not output validation (after). It uses OAP passports for identity-based policy enforcement with signed, auditable decisions. Guardrails AI and NeMo focus on LLM output quality, not tool-call authorization.
Do I need a hosted service to use APort guardrails?
No. APort supports fully local evaluation with a passport JSON file and local policy evaluator. Zero network calls. Hosted evaluation via aport.io is available for production deployments but is entirely optional.
What is the Open Agent Passport (OAP) specification?
OAP is an open standard that defines how AI agents declare their identity, capabilities, and operational limits. It includes signed decision objects for audit trails and supports both local and hosted evaluation.