LangChain is the most popular framework for building AI agents. This guide shows you how to add authorization to your LangChain agents without rewriting your application.

The Problem

LangChain agents have powerful tools. By default, there's no authorization layer between the LLM and tool execution. If the LLM decides to call a tool, it executes immediately.

Solution: Middleware

langchain_middleware.pypython

from veto import Veto, Policy
from langchain.agents import AgentExecutor
from langchain_openai import ChatOpenAI

# Initialize Veto
veto = Veto(api_key="veto_live_xxx")

# Configure policies
veto.register_policy(
    name="file_access",
    policy=Policy(
        tool="read_file",
        rules=[Policy.allow()]
    )
)

veto.register_policy(
    name="destructive_ops",
    policy=Policy(
        tool="delete_file",
        rules=[Policy.require_approval(), Policy.log_all()]
    )
)

# Apply middleware to agent
agent = AgentExecutor(
    agent=agent,
    tools=tools,
    middleware=[veto.langchain_middleware()]
)

Agent Types

Veto works with all LangChain agent types:

React agents
Plan-and-execute agents
Structured chat agents
OpenAI function agents

Full LangChain integration documentation.

LangChain Agent Authorization Guide

The Problem

Solution: Middleware

Agent Types

Related posts