AgentAssay: Token-Efficient Regression Testing for Non-Deterministic AI Agent Workflows

Researchers introduce AgentAssay, the first framework for regression testing AI agent workflows, achieving 78-100% cost reduction while maintaining statistical guarantees. The system uses behavioral fingerprinting and stochastic testing methods to detect regressions in autonomous AI agents across multiple models including GPT-5.2, Claude Sonnet 4.6, and others.

• →AgentAssay provides the first principled methodology for verifying AI agents haven't regressed after updates to prompts, tools, or models.
• →The framework achieves 78-100% cost reduction in testing while maintaining rigorous statistical guarantees through token-efficient methods.
• →Behavioral fingerprinting maps execution traces to compact vectors, enabling 86% detection power where traditional binary testing fails.
• →The system was tested across 5 major AI models and 7,605 trials, demonstrating significant improvements in regression detection.
• →Implementation includes 20,000+ lines of Python code with 751 tests and 10 framework adapters for production deployment.