Repello Ai

Repello AI develops security tooling for enterprises deploying AI systems and AI agents. Its platform is organized around what the company calls a three-phase framework: an inventory layer that discovers AI systems running in production and maintains an AI Bill of Materials (AI BOM) mapping applications, agentic workflows and exposure paths; ARTEMIS, an autonomous red teaming engine that tests AI applications and agents across text, image and audio modalities and maps findings to standards such as OWASP, NIST and MITRE; and ARGUS, a runtime security layer that applies adaptive guardrails to monitor multilingual and multimodal interactions and block malicious inputs, unsafe actions and policy violations in real time. Findings from discovery feed testing, and testing results calibrate runtime controls.

A more recent product, Workstation Lens, extends this coverage to employee endpoints. It is intended to show what coding and browser AI agents — including Claude Code, Cursor, Codex and Copilot — access on a device, which commands they run and where they send data. Described use cases include detecting SKILL.md instructions that exfiltrate data to external destinations, flagging agent access to local credential files such as ~/.aws/credentials, and identifying MCP servers whose checksums or permissions change after approval. The product can block credential access, unsafe commands and unknown outbound connections before execution, and push remediation (removal, update or blocking of unsafe configurations) across a device fleet via existing EDR and MDM tooling such as Microsoft Defender and Kandji. Repello describes an on-premises deployment posture in which files are scanned locally and only findings, not raw files, leave the device.

Repello also publishes open-source security tools through its GitHub organization, including Agent-Wiz (a CLI for threat modeling and visualizing agents built on LangGraph, AutoGen, CrewAI and similar frameworks), Whistleblower (an offensive tool for testing system prompt leakage and capability discovery of API-exposed AI applications), an MCP tool-poisoning exploit demonstration, an Artemis Python SDK, and a GitHub Action for running Artemis scans.

Founding story

Repello was established in 2024 by IIT Roorkee alumni Aryaman Behera and Naman Mishra, who framed the company around an offensive-security mindset — using attacker techniques to find and address vulnerabilities in AI systems before they are exploited.

Business model

Repello sells an enterprise software platform for AI security, marketed through demos and a 7-day pilot for Workstation Lens, with deployment through customers' existing EDR and MDM infrastructure and an on-premises option where scanning and findings remain in customer-controlled infrastructure. Specific pricing terms are not disclosed in the sources.

Not disclosed in the available sources; the company offers paid enterprise products with free entry points such as a free red teaming scan and a 7-day pilot.

Traction

Public signals include open-source traction (Agent-Wiz at 393 GitHub stars, Whistleblower at 177), customer testimonials from a head of security and from Lyzr CEO Siva Surendira, stated customers in India and among Fortune 500 companies, and inclusion in Gartner emerging-tech research on Agentic AI Security. Company-stated platform figures include 15M+ attack patterns, 270+ vulnerability types and 100+ languages supported.

Latest developments

Repello has shifted messaging toward endpoint AI agent security with Workstation Lens, which covers Claude Code, Cursor, Codex, Copilot and browser agents and integrates with EDR/MDM tools including Microsoft Defender and Kandji. It reports being named in Gartner's emerging tech report for Agentic AI Security (January 2026), published a red teaming assessment of Lyzr's AI agents (December 2025), and maintains an artemis-recon Claude Code plugin repository with commit activity into 2026.