Developers

The web app is one way to use AIScan. The same engine is available as a REST API, as an MCP server for AI assistants and as a command line tool for CI pipelines.

REST API

Every scan runs in the background. Start it, keep the returned id, then ask for the result until its status is done. No account is needed. A fair use limit per IP address applies.

# start a passive website scan
curl -X POST https://aiscan.visionedge.sk/api/scans/url \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com"}'
# answer: {"id": "SCAN_ID"}

# status, findings and verdict
curl https://aiscan.visionedge.sk/api/scans/SCAN_ID

# full report: json, sarif, md, html or cdx
curl -o report.sarif https://aiscan.visionedge.sk/api/scans/SCAN_ID/report.sarif

Code scans take a ZIP upload (/api/scans/code) or a public GitHub link (/api/scans/repo). Every endpoint with its fields is described in the interactive API reference. The machine readable version is at /api/openapi.json.

MCP server

AIScan speaks the Model Context Protocol, so an AI assistant can check a package before it installs it, or scan the code it just wrote. All tools only read; nothing is changed on your side.

Endpoint (streamable HTTP, no login): https://aiscan.visionedge.sk/mcp

Cursor and other clients with a server list

{
  "mcpServers": {
    "aiscan": { "url": "https://aiscan.visionedge.sk/mcp" }
  }
}

Codex

Add this to ~/.codex/config.toml:

[mcp_servers.aiscan]
url = "https://aiscan.visionedge.sk/mcp"

ChatGPT and Claude

Install VisionEdge AIScan with one click from the ChatGPT and Codex plugin directory, or add a custom connector with the endpoint above.

Then ask for example: "Is the npm package react-codeshift safe to install?", "Check this package.json" or "Scan https://github.com/owner/repo".

Command line and CI

The command line tool runs the same checks on your machine or in your pipeline. Your code never leaves the machine. It is in early access for now: write to info@visionedge.sk and we send you the install command.

aiscan scan .                          # code folder
aiscan scan . -f html -o report.html   # report you can share
aiscan url https://myapp.com           # live website, passive checks
aiscan feed                            # newest vulnerabilities in AI packages
aiscan mcp                             # MCP server for local agents

Exit codes for build gates: 0 nothing at or above --fail-on (default high), 1 findings at or above it, 2 error.

GitHub Actions

Findings appear under Security, Code scanning, and the job fails on critical issues.

name: AIScan
on: [push, pull_request]
jobs:
  scan:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      security-events: write
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"
      - name: Install AIScan
        run: pip install "INSTALL_SOURCE_FROM_EARLY_ACCESS_MAIL"
      - name: Scan
        run: aiscan scan . --format sarif --output aiscan.sarif --fail-on critical
      - name: Upload to GitHub code scanning
        if: always()
        uses: github/codeql-action/upload-sarif@v3
        with:
          sarif_file: aiscan.sarif

Questions or a finding that looks wrong: Support.