Tools and MCP
Pick a tool
Use the Martian MCP adapter for the lab walkthrough. Add these tools when the task needs them.
| Task | Installed tool |
|---|---|
| Drive an agent UI | Playwright CLI or Playwright MCP, with Chromium |
| Capture/replay HTTP traffic | mitmproxy: mitmproxy, mitmdump, mitmweb |
| Run curated AI regression cases | Promptfoo core |
| Build adaptive Python campaigns | PyRIT (pyrit wraps pyrit_scan) |
| Probe model responses | garak |
| Inspect MCP schemas and calls | MCP Inspector |
| Review MCP packages / skills | Cisco MCP Scanner / Skill Scanner |
| Check scoped web/API services | Nuclei, ffuf, ZAP |
| Inspect supplied source/images | Gitleaks, Trivy |
| Observe callbacks | Interactsh client/server; private collector required |
15 upstream components, 21 executable entry points. Choose one browser interface per workflow. See install and verify for maintenance.
What needs separate setup
Section titled “What needs separate setup”Evaluation targets, generators and graders need credentials/configuration. Hermes’s model connection is not inherited automatically. Interactsh’s external collector and domain are not deployed.
Cost and license
Section titled “Cost and license”Local open-source features have no required software-license fee. Top-level licenses are MIT/Apache-2.0, with an Apache/MIT transition in Inspector; dependencies and templates need separate review. Inference, AWS and optional vendor services can cost money.
Promptfoo remote-only plugins and paid Cisco analyzers are outside local coverage. Nmap, Metasploit, SQLmap, ARTEX, PentAGI and broad Kali bundles are not in this installation.