Open Source SDK Available

Secure What AI Agents Do

Not Just What They Can Access

The first real-time control plane for AI agent actions. Add one line of code to your LangChain agents and get instant visibility, policy enforcement, and human-in-the-loop approval workflows.

Get Started Free View Demo

SOC2 Ready Architecture

<10ms Policy Evaluation

100% Action Visibility

your_agent.py

from aasp_sdk import AASPCallback

# One line to secure your agent
agent = create_agent(callbacks=[AASPCallback()])

pip install aasp-sdk

AI Agents Are Autonomous. Your Security Isn't.

Classic security infrastructure was built for humans and static services. AI agents break every assumption.

AASP - AI Agent Security Platform

Control Every Agent Action

Real-time interception, policy enforcement, and audit logging for all your AI agents.

Policy Engine

Define explicit rules for what agents can and cannot do. Regex patterns, action types, allow/block/require_approval decisions.

Declarative RulesRegex Matching3 Decision Types

Action Logging

Every agent action captured with full context. Immutable audit trail for compliance and debugging.

Full ContextQueryable LogsCompliance Ready

Human-in-the-Loop

Route sensitive actions to humans for approval. Configurable workflows with expiration and escalation.

Approval QueueDashboard UISlack Coming Soon

LangChain SDK

One-line integration with LangChain. Just add our callback handler and you're protected.

1 Line IntegrationPython SDKpip install

from langchain.agents import create_react_agent
from aasp_sdk import AASPCallback

# Initialize AASP with your API key
callback = AASPCallback(api_key="aasp_live_xxx")

# Create your agent with AASP protection
agent = create_react_agent(
    llm=llm,
    tools=tools,
    callbacks=[callback]  # That's it!
)

# Every tool call is now monitored & controlled
agent.invoke({"input": "Process the invoice"})

Developer Experience

1 Line. That's It.

No complex setup, no infrastructure changes. Just add our callback handler to your existing LangChain agent and you're protected.

pip install aasp-sdk

View on GitHub PyPI Package

v0.1.0

Latest Version

<5 min

Setup Time

MIT

License

secure_agent.py

from langchain.agents import create_react_agent
from langchain_openai import ChatOpenAI
from aasp_sdk import AASPCallback

# Your existing setup
llm = ChatOpenAI(model="gpt-4")
tools = [search_tool, calculator_tool]

# Add AASP protection (1 line!)
callback = AASPCallback(
    api_key="aasp_live_xxx"
)

# Create protected agent
agent = create_react_agent(
    llm=llm,
    tools=tools,
    callbacks=[callback]
)

# Run your agent - now secured
result = agent.invoke({
    "input": "Search for quarterly revenue"
})

# Every action is:
# ✓ Logged with full context
# ✓ Evaluated against policies
# ✓ Routed for approval if needed