MCPTotal, now Autonomous Security, secures the AI agents that employees run on their laptops, in their browsers and through enterprise AI services. It is for security teams that have watched coding assistants, desktop agents and MCP servers spread across the company faster than any approval process could keep up.
The risk it addresses is that these agents act with real access: they read files, call business systems and hold tokens, often without anyone in security knowing they exist. MCPTotal gives those teams visibility of the agents in use and control over what they can reach.
At A Glance
| Vendor | MCPTotal, now Autonomous Security Inc. |
|---|---|
| Category | AI agent and MCP security |
| Identities covered | Users, AI agents, MCP servers, OAuth tokens |
| Deployment | Endpoint agent, browser extension and MCP Gateway |
| Agentic AI | MCP Gateway, token vault, MCP and skills scanner, agent guardrails |
| Certifications | SOC 2 Type II |
| Founded | 2025 |
| Market classification | Agentic AI Identity |
What MCPTotal Does
The platform follows a simple loop. It discovers which AI tools and agents are in use, assesses MCP servers, skills and other assets for risk, and lets administrators decide what is allowed. It then monitors agent behaviour as it happens, and controls how agents reach tools and systems, so that access follows the identity of each user and agent rather than shared credentials. Coverage extends to more than 60 desktop agents, including Cursor, Claude Code, Codex, ChatGPT Desktop and Gemini CLI, as well as AI used in the browser and enterprise AI SaaS.
Products
AI Security Posture
AI Security Posture answers the question most security teams start with: which AI tools and agents are people actually using, and how risky are they? It combines an AI inventory with policy management, and its MCP and skills scanner continuously checks servers and skills for vulnerabilities, malicious behaviour, risky permissions and exposed secrets.
Runtime Protection
Runtime Protection watches agent activity in real time and enforces guardrails, both model-based and deterministic, on the endpoint and in the browser. It includes data loss prevention, so prompts carrying sensitive data can be stopped before they reach an AI service.
Secure Connectivity
Secure Connectivity is for organisations that want agents to use business systems but on controlled terms. The MCP Gateway places sandboxed MCP servers behind one governed endpoint, the Token Vault keeps credentials out of the agents themselves, and access is granted by the identity and context of the user and agent.
How It Works
Setup is a reported 10-minute integration. The platform has four parts:
| Endpoint agent | A lightweight agent on macOS, Windows and Linux, installed with one command or pushed through Microsoft Intune, Jamf Pro, Iru (Kandji), JumpCloud, Fleet or another MDM. |
|---|---|
| Browser extension | Covers AI used in a browser tab, which the desktop agent cannot see, applies the same guardrails when a prompt is submitted and can block unapproved AI web apps. |
| MCP Gateway | Administrators add each MCP server once, choose which of its tools people can call, and every connection is audited. |
| Token Vault | OAuth tokens are injected only at the moment a tool call needs them, so agents, prompts and chat histories never hold the secret. |
Claude Enterprise, ChatGPT Enterprise and Microsoft 365 Copilot are covered through their own APIs.
Identities And Systems Covered
| Users and groups | Policies apply to people and teams. |
|---|---|
| AI agents | Desktop agents, browser AI such as ChatGPT, Claude, Gemini and Microsoft 365 Copilot, and enterprise AI SaaS; each connection signs in as its user. |
| MCP servers and skills | Scanned for risk and run in sandboxes. |
| Secrets | OAuth tokens in the Token Vault, and exposed secrets found by the scanner. |
| Devices | The laptops and desktops where agents run. |
Integrations
| Area | Integrations |
|---|---|
| Identity providers | Okta, Microsoft Entra ID |
| Device data | CrowdStrike Falcon, Microsoft Defender |
| SIEM and logging | Splunk, IBM QRadar, OpenTelemetry, Syslog |
| Notifications | Slack, webhooks |
Company Background
| Headquarters | New York, United States. |
|---|---|
| Leadership | Gil Dabah (Co-founder and CEO) and Dr. Ariel Shiftan (Co-founder and CTO). |
| Investors | YL Ventures and George Kurtz, Co-Founder and CEO of CrowdStrike, with backing from the founders of Snyk, Armis and Wiz. |
| Scale | More than 5 million reported agents protected in real time. |
Note: This overview is compiled from MCPTotal’s publicly available information as of 3 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.