runZero MCP — Claude Code MCP server: setup
This MCP server gives AI agents access to runZero queries, keeps a local SQLite view of the attack surface, and tracks changes over time.
Open on GitHubWhat runZero MCP does
runZero MCP exposes runZero queries to an AI agent and keeps a local SQLite copy of the full attack surface. The local data can be sliced again without using API quota, compared across time, and joined with vulnerability records while offline. Authentication with runZero is required after installation.
The package includes runzero-cli, runzero-mcp, and SKILL.md, which lets Claude Code, Codex, and Claude Cowork discover the skill. The installer supports macOS, Linux, and Windows PowerShell. MCP connections are also listed for Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, and Zed.
Use the default stdio transport. In this build, HTTP mode has no password and is exposed through network interfaces, so a web page can reach the server. That makes HTTP unsuitable even on localhost and matters before you connect a remote agent.
What it helps with
- run runZero queries through an AI agent
- keep an attack-surface snapshot in SQLite
- compare infrastructure changes over time
- join assets with vulnerabilities offline
- avoid spending API quota on repeated slicing
Example prompts for the agent
- Compare the latest attack-surface snapshots and summarize what changed.
- Find assets linked to vulnerabilities in the local SQLite data.
- Run a runZero query and group the results by asset.
How to install runZero MCP
Install
The MCP registry has no launch command for it: see the repository README
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about runZero MCP
How do I install runZero MCP in Claude Code?
Add the marketplace with /plugin marketplace add Servosity/msp-skills, then install the server with /plugin install runzero@msp-skills.
What does runZero MCP do?
It connects runZero queries to an AI agent and maintains a local SQLite copy of the attack surface. The copy supports time-based diffs and offline links between assets and vulnerabilities.
Which transport should I use with runZero MCP?
Use stdio. HTTP mode in this build has no password and can be reached through network interfaces, so it should not be run even on localhost.
