AgentGuard Python SDK

Python SDK for the AgentGuard AI Agent Authorization API.

AgentGuard provides an authorization and policy layer between AI agents and the tools or resources they interact with. It evaluates requested actions and returns one of three decisions:

• ALLOW
• BLOCK
• REQUIRE_APPROVAL

Installation

pip install agentguard

Quick Start

from agentguard import AgentGuard

client = AgentGuard(api_key="ag_live_your_api_key")

result = client.authorize(
    agent_id="sales-agent",
    action="send_email",
    resource="customer_123",
    parameters={
        "recipient": "customer@example.com"
    }
)

print(result.decision)
print(result.risk_score)
print(result.reason)

Decision Helpers

The returned EvaluationResult provides convenient properties:

if result.allowed:
    print("Action is allowed")

if result.blocked:
    print("Action is blocked")

if result.requires_approval:
    print("Human approval is required")

Approval Status

For actions requiring approval:

if result.requires_approval:
    approval = client.get_approval_status(result.approval_id)
    print(approval)

Custom API URL

The SDK defaults to the hosted AgentGuard API, but a custom API URL can be supplied:

client = AgentGuard(
    api_key="ag_live_your_api_key",
    base_url="https://your-agentguard-instance.example.com"
)

Security

Keep your AgentGuard API key secret. Do not commit API keys to source control or expose them in client-side applications.

Use environment variables or a secret manager in production.

License

AgentGuard is licensed under the MIT License. See the LICENSE file for details.