Repello AI sells security for enterprise AI systems: agentic workflows, MCP connections, chatbots and the coding agents running on employee laptops. Its products find the AI in use across an organisation, attack it the way an adversary would, and enforce guardrails while it runs.
The problem it addresses is that companies are putting AI agents and models into production faster than they can see or test them. Coding agents on developer machines add a further risk, because they handle source code, files and credentials with the developer’s own access.
At A Glance
| Vendor | Repello AI |
|---|---|
| Category | AI red teaming, runtime guardrails and AI agent security |
| Identities covered | AI agents, coding agents, models, MCP connections, credentials agents access |
| Deployment | Endpoint agent through existing EDR and MDM; API and browser based testing |
| Agentic AI | Agent red teaming, runtime guardrails, coding agent governance, MCP inventory |
| Customers | Docusign, Thomson Reuters, Groww, PhysicsWallah, Lyzr AI |
| Certifications | SOC, ISO 27001 |
| Market classification | AI Model Security; also AI Safety & Guardrails |
What Repello AI Does
Repello AI follows a discover, test and protect sequence. An organisation cannot secure AI it does not know about, so the first step is a catalogue of what is running and what each system connects to. Those systems are then attacked the way an adversary would attack them, which shows which weaknesses are exploitable and what to fix first. Finally, guardrails enforce policy while the AI runs. The same sequence applies on employee laptops, where coding agents are found, assessed and controlled on the device itself.
Products
Inventory
Inventory is for security teams that need to know what AI exists before they can protect it. It scans codebases, cloud environments and third-party apps to discover models, agents and MCP connections, maps the APIs, MCPs and plugins each one connects to, and keeps the inventory up to date as systems change.
ARTEMIS
ARTEMIS is an automated red teamer built with agentic AI in mind, chaining steps through an agent’s tools the way an attacker would. Reasoning adversaries run multi-turn attack sequences against AI applications and agents, discovering and validating multi-stage exploit chains. It tests text, image and voice inputs with a library of more than 1,000 adversarial attacks, runs in Browser Mode, through an API or inside CI/CD, and reports risk scores, evidence and mitigations mapped to MITRE, OWASP, NIST, ISO 42001 and the EU AI Act.
ARGUS
ARGUS provides real-time guardrails for LLMs and AI agents. Rules are custom and context aware across modalities, languages and domains, covering text, image and audio in more than 100 languages, with full logging, audit trails and alerts. Every agent and LLM in an organisation can be connected to it for one view of AI activity.
Workstation Lens
Workstation Lens secures the AI agents on employee devices, especially coding agents. It scans the MCP servers, skills, plugins, rules, hooks, extensions, models and memory that coding agents load, keeps evaluating them during live sessions, and allows or blocks each tool call according to the resource, command and destination involved.
How It Works
| Component | What it does |
|---|---|
| Endpoint deployment | Workstation Lens is rolled out across employee workstations through the customer’s existing EDR and MDM tools, so it sees agent activity where code, files and credentials are accessed. |
| Discovery | Inventory connects to code repositories, cloud environments and third-party AI model integrations. |
| Testing | ARTEMIS reaches AI endpoints through API integration or tests applications in a browser like a real user, and can run as a CI/CD step. |
| Runtime | ARGUS captures every call and anomaly, with rules set per application or use case. |
Identities And Systems Covered
| AI agents and models | Agents, models, datasets and tools in production |
|---|---|
| MCP connections | MCP servers, skills, plugins, hooks and extensions |
| Credentials and files | What coding agents access on the device, including credentials, files, shell commands and outbound destinations |
| Coding agents | Claude Code, Cursor, Codex, GitHub Copilot, ChatGPT Desktop, Windsurf, OpenCode, Ollama and Antigravity |
Company Background
| Leadership | Aryaman Behera, co-founder and CEO; Naman Mishra, co-founder and CTO. |
|---|---|
| Funding | A $1.2 million seed round led by Venture Highway with Pi Ventures and others; investors also include General Catalyst and Entrepreneur First. |
| Recognition | Included in Gartner research on emerging agentic AI security companies. |
Note: This overview is compiled from Repello AI’s publicly available information as of 4 November 2026. It is provided for information only. NHI Mgmt Group does not endorse any vendor, and nothing on this page represents the opinions of NHI Mgmt Group.