virgilhq 0.6.0

CLI for Virgil — security audits with the triage built in. Real scanners + clustering + LLM priority queue + code-grounded chat. Installs as `virgil` on your PATH.

virgil

Terminal client for Virgil — security audit with the triage built in. Real scanners + clustering + LLM priority queue + code-grounded chat.

The CLI is a thin shell over the Virgil service. It bundles your working directory, submits a scan, streams progress, and prints the ranked findings with CI-friendly exit codes. Scans run on Virgil's servers and bill against your account's audit credits.

Install

virgil is a CLI tool, so the right installer is pipx — it puts the binary on your $PATH and isolates the dependencies from your system Python:

pipx install virgilhq

Don't have pipx yet? Either install it once, or fall back to user-mode pip:

# macOS
brew install pipx && pipx ensurepath

# Linux / other
python3 -m pip install --user pipx && pipx ensurepath

# Fallback if you don't want pipx at all:
python3 -m pip install --user virgilhq

The PyPI package is virgilhq (the bare virgil name was already taken). The command on your $PATH is still just virgil.

Why pipx and not pip install? Modern Python distributions (Homebrew Python, Debian/Ubuntu's python3, etc.) mark the system interpreter as externally managed per PEP 668 — a bare pip install errors out. pipx quietly handles the venv for you, which is what you want for CLI tools anyway: each one gets its own isolated environment so a virgil upgrade can't break some other tool.

The CLI talks to the hosted service at https://virgilhq.app — no extra setup. Just virgil login once, then virgil scan . and you're in.

Usage

# Scan and land on triage (counts → ranked clusters → next-steps hint).
virgil scan .                                 # local directory
virgil scan                                   # …also defaults to the cwd
virgil scan github.com/OWASP/NodeGoat         # remote, bare host
virgil scan OWASP/NodeGoat                    # remote, GitHub shorthand
virgil scan https://github.com/OWASP/NodeGoat # remote, full URL

# Land on a different surface after the scan finishes.
virgil scan . --show report                   # exec narrative
virgil scan . --show surface                  # languages / frameworks / IaC profile
virgil scan . --show ask_virgil               # drop into the chat REPL pre-flighted

# PR mode — only flag findings on lines changed between two SHAs.
virgil scan . --base-sha abc1234 --head-sha def5678

# Don't wait for the scan to finish; print the audit ID and return.
virgil scan . --no-wait

# After a scan, drill in:
virgil clusters <audit-id>               # every cluster, sorted by severity
virgil clusters <audit-id> --sev high    # filter
virgil cluster  <audit-id> <key>         # one cluster in detail (prefix match ok)
virgil findings <audit-id>               # raw findings table
virgil chat     <audit-id>               # interactive Q&A grounded in this audit
virgil chat     <audit-id> -m "what's the worst finding?"   # one-shot
virgil open     <audit-id>               # launch the web app on the triage tab
virgil open     <audit-id> --page chat   # …or chat / findings / report / attack-surface
virgil status   <audit-id>

# Reports in any supported format.
virgil report <audit-id> --format md
virgil report <audit-id> --format sarif -o findings.sarif
virgil report <audit-id> --format json
virgil report <audit-id> --format pdf

Config

Persistent settings live in ~/.config/virgil/config.json:

virgil config show
virgil config set default_fail_on=high
virgil config set default_post_scan_view=ask_virgil     # triage | report | surface | ask_virgil
virgil config unset default_fail_on
virgil config path

Resolution order for each setting: env var → config file → built-in default.

CI integration

virgil scan . --fail-on critical     # exits 1 on any Critical
virgil scan . --fail-on high         # exits 1 on Critical or High
virgil scan . --fail-on never        # always exits 0

Exit codes:

Code	Meaning
0	scan finished, no findings exceeded --fail-on
1	scan finished, findings exceed the configured threshold
2	the audit itself failed (clone error, scanner crash, etc.)
3	could not reach the Virgil API

Environment

Variable	Default	What it does
VIRGIL_TOKEN	—	API token (overrides the one saved by virgil login).
VIRGIL_FAIL_ON	critical	Default --fail-on threshold for virgil scan.
VIRGIL_SHOW	triage	Default --show surface after virgil scan (triage / report / surface / ask_virgil).
VIRGIL_CONFIG_DIR	~/.config/virgil	Override the config directory.

What the output looks like

$ virgil scan .
bundle /work/myrepo → zip → submit
┌─ [ virgil ] ───────────────────────────────────────────────────────────────┐
│ audit_id  c9b1…                                                            │
│ source    scan.zip                                                         │
│ state     succeeded  phase=completed                                       │
└────────────────────────────────────────────────────────────────────────────┘

 CRIT  HIGH  MED   LOW   INFO  KEV  unreach
   2     7    14    6     3     1     19

╭─ [ fix.this_week() · ranked ] ─────────────────────────────────────────────╮
│ #01 [ CRIT ]  Hard-coded AWS access key in source  ×3                      │
│      Critical credential exposure with CISA-KEV-adjacent risk profile…     │
│ #02 [ HIGH ]  SQL injection via raw query helper  ×12                      │
│      12 callsites share src/db/query.py — fix the helper, not callsites.   │
╰────────────────────────────────────────────────────────────────────────────╯

License

Apache-2.0.