SECURITY

Exabeam launches Praxen, an open-source tool to verify AI agent behavior

Security intelligence and management solutions company Exabeam Inc. today introduced Agent Behavior Verification, a pre-deployment security discipline for artificial intelligence agents, and released Praxen, an open-source tool that tests an agent against the job it was hired to do before it ever goes live.

Praxen puts the idea into practice. The company pitches ABV as a pre-deployment check, one that runs before the runtime monitoring its Agent Behavior Analytics product already provides kicks in. Vulnerability scanning and red teaming probe an agent once it is running. ABV looks at the whole agent first and asks a blunter question: Does what it can actually do line up with the role it was given?

That question extends a push Exabeam has made through 2026. The company expanded its Agent Behavior Analytics product to cover agents across ChatGPT, Microsoft Corp.’s Copilot and Google LLC’s Gemini in April, then stretched it across Google Cloud’s agent ecosystem weeks later. ABA watches agents in production. ABV is the front end of the same strategy, applied before an agent is switched on.

Released under the Apache 2.0 license, Praxen is a reference build of the discipline. It starts from an ABV remit, which is Exabeam’s term for a policy contract setting out what an agent may do, what it may touch and where it has to stop. Praxen measures that remit against the agent’s real tools, configurations, memory, integrations and operating environment, then reports where the two diverge. Each report lists specific findings, recommends fixes and assigns a maturity score for the agent’s security posture.

“As agents become digital workers, security teams need more than runtime visibility,” said Steve Wilson, chief AI officer at Exabeam and founder and co-chair of the OWASP Gen AI Security Project. “They need confidence that agents have the right permissions, the right controls and the right boundaries before they enter production. Agent Behavior Verification helps answer a fundamental question: will this agent do its job and only its job?”

Built as an agentic coding agent skill, Praxen is aimed at developers and security practitioners who want to run the checks inside their own environments. Exabeam said it open-sourced the project to establish ABV as a shared best practice while the industry is still working out how autonomous agents should be governed and verified.

One early user pointed to the engineering output rather than the risk reporting. “The code-level remediation path it produced didn’t give us a risk report to file away,” said Medigram Inc. Chief Executive Sherri Douville. “It gave us a precise engineering roadmap we could act on immediately. In enterprise AI deployment, the gap between what an agent is authorized to do and what it is actually capable of doing is where operational risk lives.”

Praxen is available now on Exabeam’s project site.

Image: Exabeam