Verno Labs

What it is

A platform with three parts. Verno Red runs adversarial campaigns against an agent, Verno Guard watches live interactions and tool calls for attacks, and Verno Fix tracks remediations to closure. Findings map to OWASP LLM, OWASP Agentic Security, MITRE ATLAS, and CWE.

What it helps with

- Ongoing testing of text, voice, and multimodal agents
- Catching adversarial behavior in agent traffic and tool calls at runtime
- Sending findings to Jira, Slack, webhooks, or a SIEM

Who it's for

Enterprise security teams that run AI agents in production and test only agents they own or are authorized to assess.

Worth knowing

It runs as a managed service, or self-hosted in Docker on AWS, GCP, or Azure with an enterprise license.
Public-facing agents can get a free recon before a paid plan.

Can testing stay off production?
Yes. You choose whether campaigns hit production, staging, or a sandbox, and you set the rate.

Does it work with my model provider?
It lists OpenAI, Anthropic, Google, Mistral, DeepSeek, self-hosted models, LangChain, and LangGraph.

Is there an SDK?
Yes. There is a Python SDK and a command-line tool alongside the console.