The Permission Layer for the Agentic Era.
Prevent prompt injection, billing disasters, and unauthorized access with one line of code.
The general-purpose middleware server. Works with Python, CrewAI, AutoGPT, and custom loops. Centralized policy management for complex agent swarms.
Install Python SDK →
Specialized TypeScript/Node.js runtime. Features JIT token vending, @Protected decorators, and strict typing for LangChain.js and web agents.
One Architecture, Two Runtimes.
from agentsudo import AgentSudo, BudgetExceededError # 1. Initialize with your policies client = AgentSudo(agent_id="research_bot") try: # 2. Request Access (The Zero Trust Handshake) with client.request_access(tool="openai", intent="Summarize user document") as token: # 3. Execute safely response = openai.ChatCompletion.create(..., api_key=token) except BudgetExceededError: print("[-] Hourly budget limit reached.")
import { guard, Protected } from 'molt-guard'; class PaymentService { // 1. Protect methods with decorators @Protected('financial_policy') async processRefund(amount: number) { // This code ONLY runs if policy passes return stripe.refunds.create({ amount }); } } // 2. Or wrap existing tools const secureStripe = guard.protect(stripe, 'strict_mode');