APort Agent Guardrail for LangChain

APort Agent Guardrail for LangChain/LangGraph — pre-action authorization for AI agents: an AsyncCallbackHandler that verifies every tool call against your passport and policies before execution. On deny, raises GuardrailViolation.

Install

pip install aport-agent-guardrails-langchain

Requires aport-agent-guardrails (core); installs automatically.

Setup

aport-langchain setup

• Writes config to ~/.aport/langchain/config.yaml.
• Run the passport wizard with: npx @aporthq/aport-agent-guardrails --framework=langchain.

Usage

from aport_guardrails_langchain import APortCallback, GuardrailViolation
from langchain.agents import create_agent

# Add callback to your agent
agent = create_agent(model=model, tools=tools)

try:
    result = await agent.ainvoke(
        {"messages": [{"role": "user", "content": "run the task"}]},
        config={"callbacks": [APortCallback()]},
    )
except GuardrailViolation as e:
    print(f"Blocked: {e.code} - {e}")
    print("Reasons:", e.reasons)

Config is auto-loaded from .aport/config.yaml or ~/.aport/langchain/config.yaml. Override with APortCallback(config_path="/path/to/config.yaml").

Config

• mode: local | api
• passport_path: path to passport JSON (local mode)
• agent_id: ap_xxx (API mode, hosted passport)
• api_url: optional; default https://api.aport.io

Tests

cd python/langchain_adapter
pip install -e ".[dev]"
pip install -e ../aport_guardrails
pytest tests/ -v

Links

• Framework doc
• APort