mcp-orchestration

name	mcp-orchestration
description	Use when orchestrating headless tooling via MCP in this container (filesystem, fetch, playwright, semgrep, trivy, promptfoo, sqlite, kubernetes). Ensures outputs are saved under /pentest/ and avoids brittle npx cold-starts.
license	MIT
metadata	{"owner":"mayflowergmbh/kali-ai-redteam"}

MCP Orchestration (Headless)

Available MCP Servers (Configured By entrypoint.sh)

• filesystem: read/write under /pentest, /tmp, /var/log
• fetch: HTTP client (mcp-server-fetch)
• playwright: headless browser automation
• memory: scratch memory for the session
• promptfoo: run evals/redteam
• trivy: run trivy mcp for scans
• semgrep: run semgrep-mcp for repo scans
• sqlite: local db at /pentest/mcp.sqlite
• kubernetes: cluster inspection (uses kubectl config)

Conventions

• Always write artifacts to /pentest/ (never to $HOME).
• Cache expensive results and summarize them into /pentest/reports/.
• If an MCP server is unavailable, fall back to the CLI tool directly.