PolicyLayer Scan Action

GitHub Action that scans your MCP config for security risks on every PR.

Finds your MCP server configuration, analyses it against 115+ known servers, and posts a comment on the PR with a hosted report URL.

Usage

name: MCP Scan
on: [pull_request]

jobs:
  scan:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: PolicyLayer/scan-action@v1
        with:
          fail-on: high
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Inputs

Input	Description	Required	Default
fail-on	Fail the check if findings meet this severity (low, medium, high, critical)	No	(don't fail)
config-path	Path to MCP config file	No	.mcp.json

What it does

1. Reads your MCP config from the repo
2. Strips secrets (API keys, tokens, env vars, file paths are never sent)
3. Sends server identifiers to the PolicyLayer scan API
4. Posts a sticky comment on the PR with the report URL and summary stats
5. Optionally fails the check if findings exceed the severity threshold

The PR comment updates on each push (no duplicate comments).

Example PR comment

PolicyLayer MCP Scan

View Full Report

Metric	Value
Servers detected	5
Total tools	131
Dangerous tools	42

Privacy

Only server names and package identifiers are sent to the scan API. Raw config content, API keys, tokens, environment variables, and file paths are stripped before anything leaves the runner.

Links

• Scan your config online
• CLI: npx -y policylayer scan
• Intercept -- enforce limits on MCP tool calls