MCP Server

A thin adapter over the same /api/v1 endpoints, so an AI assistant can start an audit, poll it, and read findings directly in a conversation. Same auth, same rate limits, same org-scoping as every other API-key request — the MCP server has no special access of its own.

Tools

• start_scan — start an audit against an already-authorized domain.
• get_scan_status — status, depth, and severity breakdown by run id.
• get_findings — verified findings for a completed scan.
• check_gate — pass/fail against a severity threshold, same endpoint the CI/CD gate uses.
• get_report_url — a dashboard link (for a signed-in human) and an API link (for programmatic fetch).
• list_scans — your org's recent scan history.

Setup

mkdir notra-mcp && cd notra-mcp
curl -O https://your-notra-domain/notra-mcp.mjs
npm install @modelcontextprotocol/sdk zod

Then point your MCP client at it. In Claude Desktop's config:

{
  "mcpServers": {
    "notra": {
      "command": "node",
      "args": ["/absolute/path/to/notra-mcp/notra-mcp.mjs"],
      "env": {
        "NOTRA_API_KEY": "notra_live_...",
        "NOTRA_API_BASE": "https://your-notra-domain"
      }
    }
  }
}

Restart your MCP client and the six tools above become available in conversation.