agenttrust 0.1.0

Runtime security, observability, and compliance for AI agents

🛡️ AgentTrust

Runtime security, observability, and compliance for AI agents.

AgentTrust provides a comprehensive security layer for autonomous AI agents, offering real-time monitoring, threat detection, compliance enforcement, and cryptographic identity management.

Quick Start (3 lines)

import agenttrust

at = agenttrust.init(framework="langchain")
# That's it — monitoring is now active

Installation

pip install agenttrust

With LangChain integration:

pip install agenttrust[langchain]

CLI

# Initialize for your framework
agenttrust init --framework langchain

# Real-time monitoring dashboard
agenttrust monitor --live

# Check agent fleet status
agenttrust status

# View configuration
agenttrust config

Four Security Layers

Layer 1: AgentObserve

Traces every agent action with <5ms overhead. Tool calls, LLM invocations, chain executions — all captured.

Layer 2: AgentProtect

Behavioral firewall with pattern-based threat detection: prompt injection, privilege escalation, data exfiltration, and behavioral drift.

Layer 3: AgentComply

Compliance engine supporting EU AI Act, NIST RMF, ISO 42001, and SOC2. Automated audit trails and risk dashboards.

Layer 4: AgentIdentity

Cryptographic identity management for agents. Register, verify, and compute trust scores across your agent fleet.

License

MIT