Your first scan

Every Pencheff scan runs the full Deep assessment — every module, Pulse, and attack chains, mapped to OWASP Top 10 and the compliance frameworks. Pick the surface you want to work from:

Open app.pencheff.com and sign in.
Register target → enter the URL, optional credentials, scope, exclude paths.
Click Start scan — every scan runs the full Deep assessment (no profile to pick), and you’ll be redirected to a live progress page with SSE-streamed stages.
When it finishes, browse findings by severity, OWASP category, compliance framework.

> Pentest https://juice-shop.herokuapp.com, report to slack

I'll run a full Deep scan now.

An MCP-driven host (Cursor, Continue, Cline, Zed, …) uses the MCP tools in this order (fully automatic):

pentest_init → new session
check_dependencies → see which external tools are available
recon_passive → recon_active → recon_api_discovery
scan_waf → payload_generate
scan_injection → test_endpoint (manual exploitation)
scan_client_side → scan_auth → scan_authz → scan_advanced
exploit_chain_suggest → test_chain
generate_report and export_to_slack

The agent will pause to explain each finding and prove exploitability with a PoC before reporting it.

# Headless single-target scan — outputs JSON + Word report.
# Every scan runs the full Deep assessment; --profile is optional.
pencheff scan \
  --target https://juice-shop.herokuapp.com \
  --format json,docx \
  --output ./reports/
 
# Or run a YAML policy:
pencheff run-policy policies/examples/owasp_top10.yaml
 
# CI/CD gate — non-zero exit on any HIGH+ finding.
pencheff scan \
  --target $TARGET_URL \
  --fail-on high \
  --format json \
  --output ./reports/

LLM endpoints are a third target kind alongside URL and Repo. Register the chat-completions URL once, paste your API key as an Authorization: Bearer … header, pick a red-team profile, and Pencheff fires all 10 OWASP LLM Top 10 (2025) modules at it.

pencheff llm-redteam \
  --target https://your-llm.example.com/v1/chat/completions \
  --provider openai-chat \
  --model your-model-id \
  --header "Authorization=Bearer your-token" \
  --strategies base64,jailbreak,crescendo,leetspeak \
  --datasets donotanswer,harmbench \
  --guardrails pii,secrets,unsafe-code,tool-authz \
  --judge-provider openai-moderation \
  --judge-endpoint https://your-judge.example.com/moderations \
  --max-rps 0.3 \
  --max-cost-usd 5 \
  --output-format html \
  --output-file llm-report.html \
  --fail-on high

In the SaaS UI:

Register target → pick LLM endpoint.
Endpoint URL = the chat-completions URL (not the model info page — a common mistake).
Add an Authorization header row with the value Bearer sk-… (the literal word Bearer plus a space, then your key).
Pick provider and model — every red-team scan runs the full library.
Optionally configure a judge (openai-moderation is recommended for reasoning-model targets), an attacker LLM for PAIR, an embedder for semantic verdicts, and rate limits (default Max RPM = 18 to stay under OpenRouter’s free-tier cap).

Findings are mapped to OWASP LLM Top 10, MITRE ATLAS, NIST AI RMF, and EU AI Act. Compare two scans at /scans/compare?a=…&b=… to gate PRs on safety regressions.

Pencheff will never send traffic outside the scope you declared. Targets are private by default, and credentials are Fernet-encrypted at rest in the hosted SaaS, or MaskedSecret-wrapped (never logged) in the CLI.

Scan depth

There’s no profile to pick — every security scan runs the full Deep assessment: every module + Pulse + attack chains, full swarm orchestration (Tier 2 · all 7 phases · top-1000 ports · subdomain fan-out ≤100), deterministic orchestrator + MITRE ATT&CK narrative, and PCI-DSS / SOC 2 / ISO 27001 / NIST / HIPAA mappings. Expect ~60+ min. LLM red-team scans likewise always run the full payload library.

The API and CLI still accept a profile field (quick, standard, deep, plus the legacy specialised names) for backward compatibility, but every non-load value is coerced to deep at scan time — new integrations can simply omit it. The one exception is profile: "load", a separate opt-in load/performance test on live web/API targets (not a security scan). list_scan_profiles still enumerates the built-in module sets, but note that every scan runs Deep.

What’s next

Core concepts — understand sessions, findings, chains
Features — deep-dive each scan area
LLM red team — adversarial testing for chat endpoints
Repo scanning — connect a GitHub repo and scan source + deps

Installation Core concepts