prismor-cli

Prismor is an AI agent security control plane for enterprises. It is built to keep agents from becoming rogue by intercepting tool calls before they execute, adding security guardrails in production, and preserving traceability for each action.

The product watches agent activity tool by tool. It can block destructive calls, strip secrets before they reach the model, and record every call in an audit trail that can be reviewed later. The interface shown in the product includes verdicts such as allowed, blocked, and masked, along with a live session timeline and a breakdown of activity by environment, including production, staging, local, and CI. It also presents agent activity across sessions and flags calls such as credential reads, secret exfiltration, and destructive actions.

Prismor works with the OpenAI Agents SDK, MCP agents like Claude Code and Cursor, LangChain, and CrewAI. The page says it can be set up locally with a CLI, and it describes policy management as code through the CLI, where teams can write rules once, apply them, set organization defaults, override them per team, and grant per-person exemptions. It also mentions access permission profiles for least privilege and a dashboard that shows what happened before and during execution.

The service is described as used by developers, and it is also framed as a product for enterprises. No pricing or license details are stated in the provided material.