vigolium

Vigolium - High-fidelity vulnerability scanner fusing agentic AI with native speed, modularity, and precision

Vigolium provides two complementary scanning modes:

• Native Scan (vigolium scan): Fast, powerful, and flexible. Deterministic, multi-phase scanning with 317 modules across content discovery, browser/SPA spidering, and active/passive audit, covering injection, access control, file/path, API/protocol, framework-specific, cloud/infra, and out-of-band (OAST) vulnerability classes.

• Agentic Scan (vigolium agent): Thoroughly audits your codebase. AI-driven scanning that autonomously plans attacks, selects modules, generates custom extensions, and triages results, combining deep source-code audit with autonomous and targeted vulnerability scanning.

Installation

Quick Install (Recommended)

curl -fsSL https://vigolium.com/install.sh | bash

npm

npm install -g @vigolium/vigolium

Windows

The npm install above works on Windows. Alternatively, download vigolium_<version>_windows_amd64.zip from the releases page, extract it, and put vigolium.exe somewhere on your PATH.

Windows ships as x64 only; on Windows ARM it runs under emulation. The shell installer above is POSIX-only; a zip install upgrades in place with vigolium update, and an npm install by re-running the npm install.

docker pull j3ssie/vigolium:latest
docker run --rm j3ssie/vigolium:latest scan -h

git clone https://github.com/vigolium/vigolium.git
cd vigolium
make build         # build and install to $GOPATH/bin

UI Dashboard	Traffic Dashboard

Static Reports	Static Reports

Native scan	Agentic Scan

Sponsor

_{Thank you to Daytona for sponsoring the sandbox infrastructure}

Key Features

Native Scan

• 331 scanner modules: 213 active (fuzzing) + 118 passive (pattern matching), covering OWASP Top 10 and beyond
• Out-of-band testing (OAST): blind XSS/SSRF/command injection via interactsh callbacks with automatic payload correlation
• Value-aware mutation: classifies parameters by semantic type (integer, UUID, JWT, email) and mutates per intent
• Multi-phase pipeline: external harvesting, content discovery (Deparos), browser/SPA spidering (Spitolas), and audit, controlled by strategy presets and scanning profiles
• Flexible inputs: URLs, OpenAPI/Swagger, Postman, Burp Suite, cURL, Nuclei JSONL
• Multi-session authentication: inline sessions, session files, or full auth configs with login flows, token extraction, and IDOR/BOLA testing
• JavaScript extensions: custom modules and hooks via embedded JS engine with session-aware HTTP APIs
• Scalable & reportable: concurrent worker pool with per-host rate limiting, hybrid in-memory/disk/Redis queue, and self-contained HTML reports

Agentic Scan

• In-process olium runtime: every agent mode runs on the native Go pkg/olium engine: turn-based loop, built-in tool registry, skills support, and pluggable provider drivers (no subprocess SDK pools)
• Autopilot: agent autonomously discovers endpoints, runs scans, and triages findings, with optional multi-specialist pipeline and session resume
• Swarm: master agent selects modules, generates custom JS attack extensions, runs code audit + SAST, executes scans, and triages results; targeted or full-scope (--discover), with --diff/--last-commits for change-focused runs
• Source-audit drivers: audit, piolium, and the unified audit dispatcher run foreground source-code audits sharing one finding schema and DB tagging
• Query mode: single-shot prompts for code review, endpoint discovery, and secret detection
• Pluggable providers (auto-detected; set agent.olium.provider to choose): openai-compatible, openai-codex-oauth, openai-api-key, openai-responses, anthropic-api-key, anthropic-oauth, anthropic-cli, anthropic-compatible, anthropic-vertex, google-vertex. Same modes exposed over the REST API with SSE streaming and an OpenAI-compatible chat endpoint

Quick Start: Native Scan

# Scan a single target (default: balanced strategy)
vigolium scan -t https://example.com

# Scan with a strategy preset
vigolium scan -t https://example.com --strategy deep

# Scan specific modules only
vigolium scan -t https://example.com -m xss-reflected,sqli-error

# Scan from an OpenAPI spec
vigolium scan -i openapi.yaml -I openapi -t https://api.example.com

# Pipe URLs from stdin
cat urls.txt | vigolium scan

# Run a single phase directly
vigolium run discovery -t https://example.com

# Generate an HTML report
vigolium scan -t https://example.com --only discovery --format html -o report.html

See the architecture overview for the full pipeline and the strategies guide for strategies, profiles, and pace configuration. For a quick command reference, see docs.vigolium.com/getting-started/cheat-sheet.

Server Mode

# Start API server with authentication (key from config server.auth_api_key,
# or override it with VIGOLIUM_API_KEY)
VIGOLIUM_API_KEY=my-secret-key vigolium server

# Enable transparent HTTP proxy for traffic recording
vigolium server --ingest-proxy-port 9003

# Auto-scan ingested traffic (what gets scanned is limited by the scope.* rules)
vigolium server --scan-on-receive

# Ingest traffic to a running server
cat urls.txt | vigolium ingest -s http://localhost:9002

# Ingest an OpenAPI spec
vigolium ingest -s http://localhost:9002 -i api.yaml -I openapi

See running the server for server setup, ingestion for ingestion workflows, and the API overview for the full REST API reference.

Proxy integrations: forward live proxy traffic to a running Vigolium server with the burp-vigolium extension for Burp Suite or the caido-vigolium plugin for Caido. Both serve the same bridge protocol (-B/--burp-bridge-url, alias --caido-bridge-url), and ingested traffic is labelled with the proxy it came from.

Authenticated Scanning

Vigolium supports multi-session authenticated scanning for IDOR/BOLA testing and privilege escalation checks:

# Inline session via CLI flag (name:Header:value)
vigolium scan -t https://example.com \
  --auth "admin:Cookie:session_id=abc123" \
  --auth "user:Cookie:session_id=xyz789"

# Load session(s) from a YAML/JSON file
vigolium scan -t https://example.com --auth-file ./admin-session.yaml

# Auth file with an automated login flow (token extraction, etc.)
vigolium scan -t https://example.com --auth-file ./login-flow.yaml

# Add custom headers (works with sessions)
vigolium scan -t https://example.com -H "Authorization: Bearer token123"

Auth files support static headers, bearer tokens, and automated login flows with token extraction from cookies, JSON responses, or headers. Preset examples are available in public/presets/sessions/. See the authentication guide for the full guide.

The --auth / --auth-file flags were previously named --session / --session-file. The old names still work as deprecated aliases.

Agentic Scan

AI-driven scanning where agents autonomously plan, execute, and triage vulnerability assessments with the native scan engine underneath:

# Autopilot: autonomous AI-driven scanning (in-process olium engine)
vigolium agent autopilot -t https://example.com
vigolium agent autopilot -t https://example.com --source ./src --prompt "focus on auth bypass"
vigolium agent autopilot -t https://example.com --diff main...feature/auth   # diff-focused
vigolium agent autopilot -t https://example.com --intensity deep             # preset bundle

# Swarm: AI-guided targeted or full-scope vulnerability scanning
vigolium agent swarm -t https://example.com/api/users --vuln-type sqli
vigolium agent swarm -t https://example.com --discover                       # full-scope
vigolium agent swarm -t https://example.com --source ./src --discover        # source-aware full-scope
vigolium agent swarm --input "curl -X POST https://example.com/api/login -d '{\"user\":\"admin\"}'"

# Source-audit drivers (separate harness, do not route through olium)
vigolium agent audit --source ./src                                  # default: auto (audit, fall back to piolium)
vigolium agent audit --source ./src --driver audit --mode deep       # vigolium-audit only (claude/codex)
vigolium agent audit --source ./src --driver piolium --mode balanced # Pi-native (pi extension) only
vigolium agent audit --source ./src --driver both                    # audit then piolium, back-to-back
vigolium agent audit --source ./src --modes deep,confirm             # chain modes (same as --intensity deep)
vigolium agent audit --source ./src -S --output-dir ./audit-out      # throwaway DB + bundled HTML report
vigolium audit --source ./src                                        # top-level alias

# Direct olium access (TUI or headless)
vigolium ol                             # launch the olium TUI
vigolium ol --prompt "..."              # one-shot prompt (-p implies headless)

Agentic scan modes:

• Autopilot: autonomous scanning. CLI calls pkg/olium/autopilot.Run directly; the server adds vigolium-audit prep, auth setup, and a frozen context bundle around the same loop
• Swarm: AI-guided vulnerability scanning supporting targeted single-request and full-scope (--discover). Master agent analyzes inputs, selects modules, generates custom JS extensions, runs code audit and SAST, executes scans, and triages results
• Audit: source-code audit via vigolium agent audit — a unified dispatcher that runs the embedded vigolium-audit (claude/codex) and/or piolium (Pi-native) harnesses, selected with --driver {auto|both|audit|piolium} (default auto: preflight the audit leg and run it when the resolved claude/codex CLI is on PATH, the binary is embedded, and the chain has an audit-supported mode; otherwise fall back to piolium without launching audit. A mid-run audit failure surfaces rather than switching drivers). Separate harnesses; do not route through olium. Per-driver child rows under one parent AgenticScan with post-pass findings dedup. There is no standalone agent piolium subcommand — piolium runs via --driver=piolium

Standalone audit CLIs: the agentic security audit also ships as standalone CLIs you can run independently of Vigolium: vigolium-audit (the harness behind vigolium agent audit) and piolium (the Pi-native driver behind vigolium agent audit --driver=piolium).

See the agent mode guide for the full guide.

⚡ Vigolium Cloud Console

A cloud-based solution for teams that want the power of Vigolium without managing infrastructure. Console is the upgraded, fully-featured version of Vigolium, led by an advanced autonomous agent that drives the whole engagement — reasoning about the target, choosing its own attack paths, and confirming what it finds — with hosted scanning and shared results layered on top of the open-source core, so you can focus on fixing vulnerabilities instead of maintaining tooling.

Interested? Get in touch for a demo.

Native Scan Layers

The native scan pipeline is composed of modular layers, each documented separately:

Documentation

Full documentation lives at docs.vigolium.com. Release notes and version history are in the CHANGELOG. Quick links:

JavaScript Engine

Run JavaScript/TypeScript code directly or write custom scan modules and hooks without recompiling:

# Execute inline JavaScript
vigolium js --code 'let r = vigolium.http.get(TARGET); console.log(r.status)' --target https://example.com

# Run a JS file with timeout
vigolium js --code-file ./my-script.js --target https://example.com --timeout 60s

# Manage extensions
vigolium ext ls                # list loaded extensions
vigolium ext docs --example    # browse API with code examples
vigolium ext preset            # install starter scripts

The JS engine exposes session-aware HTTP APIs for authenticated testing:

// Create a persistent session with shared cookie jar.
// post() takes a string body — serialize objects yourself.
let session = vigolium.http.session();
session.post(
  "https://app.example.com/login",
  JSON.stringify({ user: "admin", pass: "secret" }),
  { headers: { "Content-Type": "application/json" } }
);
session.get("https://app.example.com/dashboard"); // cookies auto-sent

// Automated login flow with token extraction
let authed = vigolium.http.login({
  url: "https://app.example.com/api/auth",
  method: "POST",
  body: JSON.stringify({ username: "admin", password: "pass" }),
  extract: [{ source: "json", path: "$.token", apply_as: "Authorization: Bearer {value}" }]
});

// IDOR/BOLA testing across multiple sessions
let results = vigolium.http.authTest({
  sessions: { admin: adminSession, user: userSession },
  requests: [{ method: "GET", url: "https://app.example.com/api/users/1" }]
});

// Multi-step authentication sequences
let result = vigolium.http.sequence([
  { url: "/csrf", extract: [{ source: "cookie", name: "csrf_token", as: "token" }] },
  { url: "/login", method: "POST", body: "csrf={token}&user=admin" }
]);

// Parallel request batching (race conditions, IDOR)
let responses = vigolium.http.batch([req1, req2, req3], { concurrency: 10 });

// CSRF token extraction
let csrf = vigolium.http.csrf("https://app.example.com/form");

// HTTP request replay with variations
let varied = vigolium.http.replay(rawRequest, [
  { headers: { "Authorization": "Bearer admin_token" } },
  { headers: { "Authorization": "Bearer user_token" } }
]);

See writing extensions for the extension authoring guide and pkg/jsext/vigolium.d.ts for the full TypeScript API definitions.

CLI Reference

Scanning:
  vigolium scan                Run a native scan (deterministic multi-phase vulnerability scanning)
  vigolium run <phase>         Run a single native scan phase (alias for scan --only <phase>)
  vigolium scan-url <url>      Quick native scan of a single URL
  vigolium scan-request        Native scan from a raw HTTP request

Agentic scan (in-process olium engine):
  vigolium agent autopilot     Autonomous AI-driven vulnerability scanning
  vigolium agent swarm         AI-guided targeted or full-scope vulnerability scanning
  vigolium agent query         Single-shot prompt (code review, endpoint discovery)
  vigolium agent olium         Direct olium TUI (or one-shot non-interactive via -p)
  vigolium agent audit         Unified driver dispatcher (vigolium-audit and/or piolium, --driver=auto|both|audit|piolium)
  vigolium agent session       Browse/replay agent session artifacts
  vigolium olium | vigolium ol Top-level alias for `vigolium agent olium`

Server & ingestion:
  vigolium server              Start the API server with traffic ingestion
  vigolium ingest              Ingest traffic to a running server
  vigolium storage             Interact with cloud object storage (uploads, downloads)

Data & projects:
  vigolium db                  Database operations (list, stats, export, clean, seed)
  vigolium finding             Browse and manage findings (load, tui)
  vigolium traffic             Browse and replay HTTP records (tui, replay)
  vigolium replay              Mutate a stored/supplied HTTP request and diff baseline vs replay
  vigolium project             Manage projects (create, list, use, config)
  vigolium scope               Manage scope rules
  vigolium import              Import findings/data from external sources
  vigolium export              Export scan results

Extensions & auth:
  vigolium js                  Execute JavaScript/TypeScript code
  vigolium ext                 Manage JavaScript extensions (eval, lint)
  vigolium auth                Manage authentication sessions (list, load, lint, totp)

Setup & introspection:
  vigolium init                Initialize a Vigolium workspace
  vigolium config              Manage configuration (ls, set, path, clean)
  vigolium strategy            Inspect scanning strategies and phases
  vigolium module              Inspect/enable scanner modules
  vigolium doctor              Diagnose environment & dependencies
  vigolium version             Show version info

Native Scan (vigolium scan / run):
  -t, --target           Target URL
  -T, --target-file      File containing target URLs
  -i, --input            Input file path (- for stdin)
  -I, --input-mode       Input format: urls, openapi, swagger, burp, curl, nuclei, har
  -m, --modules          Modules to run (comma-separated or 'all')
      --strategy         Strategy preset: lite, balanced, deep
      --scanning-profile Scanning profile name or YAML path
      --only             Phases to run (comma-separated): ingestion, discovery (deparos),
                         external-harvest, spidering (spitolas), known-issue-scan,
                         dynamic-assessment, extension
      --skip             Phases to skip (repeatable, same names as --only)
  -S, --stateless        Use a throwaway temp database, discarded after the scan
      --fail-on          Exit non-zero when a finding at/above this severity is present

Authentication:
      --auth              Inline session definition (name:Header:value, repeatable)
      --auth-file         Session YAML/JSON file path, supports login flows (repeatable)
  -H, --header           Custom HTTP header (repeatable)

Performance:
  -c, --concurrency      Concurrent workers (default: 25)
  -r, --rate-limit       Max requests/sec (default: 100; 0 = unlimited)
      --max-per-host     Per-host concurrency cap (default: 40)
      --proxy            HTTP/SOCKS5 proxy URL
      --timeout          HTTP request timeout (default: 15s)

Agentic Scan (vigolium agent autopilot / swarm / query):
      --source             Path to source code for source-aware scanning
      --files              Specific files to include relative to --source
      --source-label       Label for source code ingestion
      --provider           Olium provider (auto-detected; set agent.olium.provider):
                           openai-compatible, openai-codex-oauth,
                           openai-api-key, openai-responses, anthropic-api-key,
                           anthropic-oauth, anthropic-cli, anthropic-compatible,
                           anthropic-claude-sdk-bridge, anthropic-vertex, google-vertex
      --model              Model ID override
      --oauth-token        OAuth bearer token (anthropic-oauth)
      --oauth-cred         OAuth/SA file path (openai-codex-oauth, anthropic-vertex,
                           google-vertex)
      --llm-api-key        API key (anthropic-api-key, openai-api-key)
      --vuln-type          Vulnerability type focus (sqli, xss, ssrf, ...)
      --prompt             Free-text task guidance (same as the positional [prompt])
      --plan-file          Plan file mixing guidance + raw seed HTTP request(s)
      --knowledge-base     File/dir describing the app; prose is distilled, traffic
                           exports (HAR/Burp/curl/OpenAPI/Postman) are ingested
      --prior-context      Front-load existing project traffic/findings: auto, summary, off
      --intensity          Preset bundle: quick, balanced, deep
      --diff               Diff range / PR URL / HEAD~N for change-focused scans
      --last-commits       Shorthand for --diff HEAD~N
      --code-audit         Enable AI code audit (default: on with --source)
      --discover           Run discovery+spidering before planning (swarm)
      --audit              Nested vigolium-audit harness: lite, balanced, deep, mock, off
                           (default off; needs --source)
      --piolium            Nested piolium harness mode (default off; needs --source)
      --resume             Resume a durable-autopilot run by agentic-scan UUID
      --session-dir        Pin the session dir for this run's debug artifacts
      --transcript         Copy transcript.jsonl out after the run
      --max-iterations     Max triage-rescan iterations
      --max-commands       Cap on agent tool calls (autopilot / swarm)
      --token-budget       Cap on input+output tokens (autopilot / swarm)
      --max-cost-usd       Cap on estimated spend in USD (autopilot / swarm)
      --max-duration       Max agent wall-clock time (0 = no limit)
                           (a run stopped by any budget exits 5)
      --preflight-timeout  Provider preflight deadline (default 15s; 60s local)
      --no-preflight       Skip the provider preflight ping
      --only / --skip / --start-from   Phase control (swarm)

Source audit (vigolium agent audit / vigolium audit):
      --source             Required: dir (--source . for cwd), git URL, gs://, archive
      --driver             auto (default), both, audit, piolium
      --intensity          Preset: quick, balanced, deep (deep = modes deep,confirm)
      --mode               Mode override: lite, balanced, deep, revisit, confirm, merge, ...
      --modes              Chain modes back-to-back (e.g. deep,confirm)
      --list-modes         Print the audit mode graph and exit
      --agent              Coding agent for the audit leg: claude or codex
      --keep-raw           Keep raw output under <source>/vigolium-results/ (on by default)
      --clean-raw          Remove the source-tree raw copy after the run
  -S, --stateless          Run into a throwaway DB and auto-render an HTML report
      --output-dir         Bundle the HTML report + raw results into one folder (needs -S)
      --no-dedup           Skip the post-pass project-wide findings dedup
      --no-preflight       Skip the pre-audit auth/model roundtrip checks
  -i, --interactive        Drive the audit yourself in the coding agent (audit driver only)

JavaScript:
      --code             Inline JavaScript to execute
      --code-file        Path to JS/TS file to execute
      --timeout          Execution timeout (default: 30s)

Output:
  -j, --json             Compact, token-aware JSON output (read/query commands)
      --format           Output format (comma-separated for multiple): console, jsonl,
                         html, sqlite (needs -S), fs (flat traffic/finding tree)
  -o, --output           Output file path
      --silent           Suppress all output except findings
  -v, --verbose          Verbose logging

Repository Layout

The platform/ directory contains external tooling, UI Dashboard and is not part of the core scanner. No changes should be made to it.

Benchmarks

Vigolium is continuously benchmarked against intentionally vulnerable applications and also heavily tested against real-world targets through bug bounty and responsible disclosure programs.

• Self-hosted (Docker): DVWA, OWASP Juice Shop, VAmPI, crAPI, Vulnerable Java App, Vulnerable Nginx, OopsSec Store (custom Next.js app)
• External (hosted): Acunetix TestPHP, Gin & Juice Shop, Testfire
• XSS & multi-vuln: BruteLogic XSS, XBOW (XSS, SQLi, SSTI, LFI, SSRF, XXE, command injection)

Run benchmarks with make test-canary (Docker apps) or make test-integration (XSS).

Development

make build          # build and install
make test           # run all tests (auto-installs gotestsum)
make test-unit      # fast unit tests (-short, no external deps)
make test-e2e       # E2E tests (requires Docker)
make lint           # run linter
make fmt            # format code

See HACKING.md for the full build guide, codebase map, and module development guide.

Security

Vigolium is an offensive security tool, and two parts of it are intentionally permissive: agent mode runs with no sandbox (the LLM has full shell, file, and network access on the host) and extensions can run arbitrary commands. Run agent mode in a disposable container/VM scoped to the engagement, and treat untrusted extensions like untrusted code. See SECURITY.md before you start, and report vulnerabilities in Vigolium itself privately to contact@vigolium.com.

License

Vigolium is released under the MIT License.

Crafted with ♥ by @j3ssie, with @theblackturtle as a core initial contributor.