MCP
Connect an AI assistant to Barrion, and control what it can spend and test from Settings.
The Barrion MCP server lets an assistant such as Claude, Cursor or Codex start scans and pentests and read what they found. It signs in as you, so it can do what you can do in the dashboard and no more.
Setup, supported clients and the full tool list are on barrion.io/mcp.
Quickstart
Add this to your client's MCP config:
{
"mcpServers": {
"barrion": {
"command": "npx",
"args": ["-y", "@barrion/mcp"]
}
}
}In Claude Code, run claude mcp add barrion -- npx -y @barrion/mcp instead.
The first tool call opens a browser to sign in to Barrion. It needs Node.js 20.18.1 or newer.
Settings
Once an assistant is connected, open Settings, then MCP in the dashboard to control it.
- Budget: the most credits assistants can spend on pentests in one billing cycle. Pentests started from the dashboard or with an API key do not count toward it. Set a budget before an assistant can start a pentest.
- Targets: which of your verified targets an assistant may pentest. This is separate from the API runs switch under API keys, and it is on by default for verified targets. When it is off for a target, assistants can only read results for it.
- Pentests started through MCP: each run with its target, level, status and credits.
- Connect another assistant: the same config snippets as above.
In an organization, your role decides what you can change here.