Skip to main content
Upload qai SARIF output to GitHub Code Scanning to surface MCP server findings in the Security tab, pull request checks, and security alerts.

​
Prerequisites

  • GitHub repository with Actions enabled
  • GitHub Advanced Security (required for private repos; free for public repos)
  • qai v0.5.1 or later

​
Basic Workflow

Create .github/workflows/qai-security.yml: 
name: MCP Security Audit

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

permissions:
  contents: read
  security-events: write

jobs:
  qai-scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.12'

      - name: Install qai
        run: pip install q-uestionable-ai

      - name: Run MCP server audit
        run: |
          qai audit scan \
            --transport stdio \
            --command "npx @modelcontextprotocol/server-memory" \
            --format sarif \
            --output results.sarif
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

      - name: Upload SARIF to GitHub Security
        uses: github/codeql-action/upload-sarif@v3
        with:
          sarif_file: results.sarif
          category: qai-mcp-audit
        if: always()

​
How It Works

  1. The workflow installs qai and runs audit scan with --format sarif
  2. The SARIF file contains findings mapped to rules (one per scanner category) with severity levels
  3. upload-sarif pushes results to GitHub Code Scanning
  4. Findings appear in the repository’s Security > Code scanning alerts tab

​
What Shows Up in GitHub

Each qai finding becomes a Code Scanning alert with:
  • Rule ID — Scanner rule identifier (e.g., QAI-INJ-CWE-078-shell_injection)
  • Severity — Mapped from qai severity: CRITICAL/HIGH → error, MEDIUM → warning, LOW/INFO → note
  • Security severity score — Numeric score for GitHub’s severity ranking (9.0 for critical, 7.0 for high, 4.0 for medium)
  • Description — Full finding description with evidence
  • Properties — Category, tool name, evidence, remediation, and mitigation data

​
Scanning Specific Checks

Use --checks to run only specific scanners in CI for faster, focused scans: 
- name: Run injection checks only
  run: |
    qai audit scan \
      --transport stdio \
      --command "npx @modelcontextprotocol/server-memory" \
      --checks injection,auth,token_exposure \
      --format sarif \
      --output results.sarif

​
Branch Protection

Block merges when high-severity findings are detected:
  1. Go to Settings > Branches > Branch protection rules
  2. Enable Require status checks to pass before merging
  3. Add the MCP Security Audit job as a required check
The upload-sarif action does not fail the workflow based on alert severity — it only errors on upload or validation failures. To block merges on findings, add an explicit CI gate step in the MCP Security Audit job that parses the SARIF output and exits non-zero when findings exceed your severity/threshold policy. Branch protection requires the named status check to pass, but blocking on finding severity requires this custom gate.

​
Managing Alerts

In the Security tab, each alert supports:
  • Open — Active finding, needs attention
  • Fixed — Resolved in a subsequent scan (GitHub detects automatically)
  • Dismissed — Manually closed with a reason (false positive, won’t fix, used in tests)
Dismissed alerts won’t reappear on future scans unless the dismissal reason is changed.

​
Troubleshooting

Results not appearing: Verify the workflow has security-events: write permission. Check the Actions log for upload errors. Allow ~30 seconds for the Security tab to update after upload. SARIF validation errors: Validate the file before upload: python -m json.tool results.sarif > /dev/null
GitHub has a 100MB SARIF file size limit. For large scans, filter with --checks to reduce output size.

​
See Also