AgentAssert

LIVE

Behavioral Contracts for AI Agents

AgentAssert brings formal behavioral contracts to AI agents. Define preconditions, postconditions, and invariants using a domain-specific language. The system detects behavioral drift using Jensen-Shannon divergence and automatically triggers recovery re-prompting when agents violate their contracts.

$pip install agentassert

Read the Paper →GitHub →Website →

Domains Validated

Models Tested

18K

Sessions Analyzed

Pillars

Features

ContractSpec DSL

A domain-specific language for defining behavioral contracts: preconditions, postconditions, invariants.

Drift Detection

Jensen-Shannon divergence monitoring. Detect when agent behavior deviates from contracted norms.

Recovery Re-Prompting

Automatic intervention when contracts are violated. Re-prompt agents back to compliant behavior.

(p,δ,k)-Satisfaction

Probabilistic contract satisfaction with confidence bounds. Statistical guarantees for stochastic agents.

OU Drift Dynamics

Ornstein-Uhlenbeck process modeling of behavioral drift with Lyapunov stability proofs.

Compositional Guarantees

Contracts compose across multi-agent systems. Verify pipeline-level compliance from component contracts.

Use Cases

Compliance verification for enterprise agents

Multi-agent behavioral contracts

Drift detection in production agents

Regulatory compliance (EU AI Act)

Research Paper

AgentAssert: Behavioral Contracts for AI Agent Compliance

Varun Pratap Bhardwaj, 2026

Read on arXiv→

Licensed under Elastic License 2.0