The ARMO MCP server lets you ask your AI assistant about your ARMO security data in plain language. It connects Claude, ChatGPT, Cursor, VS Code and any other Model Context Protocol client to your ARMO account.

Your assistant can read findings across vulnerabilities, container images, Kubernetes posture controls, cloud posture, cloud compliance, cloud accounts, runtime incidents, risks, attack chains, inventory and repositories. It can also list the ARMO accounts you have access to. Every tool is read-only: the server can't start a scan, change a setting or dismiss a finding.

ARMO hosts the server, so there is nothing to install. You point your assistant at a URL and sign in with your ARMO account. Sign-in uses standard OAuth 2.1, and your assistant registers itself with ARMO the first time it connects, so you don't need a client ID, client secret or API key.

Connect Your Assistant

Setup differs by assistant. Follow the page for yours:

Claude Desktop, claude.ai and Claude Code
ChatGPT
Cursor
VS Code
Any other MCP client

To sign in you will need The server URL for your sign in region - as mentioned below:

cloud.armosec.io = https://api.armosec.io/mcp
cloud.us.armosec.io =https://api.us.armosec.io/mcp

Sign- in & Permissions

You sign in with your ARMO account the first time your assistant connects. Sign-in uses OAuth 2.1, so no access key is needed.

Your assistant sees only what your ARMO role lets you see in the dashboard. Each colleague who connects sees their own view.
A session lasts about an hour. After that, your assistant asks you to sign in again.
Query results aren't cached, so revoking your session takes effect immediately.
Your questions go to your AI assistant's provider. ARMO sees only the tool calls the assistant makes.