sast — by Insomnia

Free, fast static application security testing for your terminal & CI/CD.
SAST + taint · secrets with live key validation · SCA/CVEs · IaC · CMS & web-shell/malware — one self-contained binary, the same engine in your IDE and your build.

Get the editor plugins — same engine, same results

Plugin home · CI/CD & SDLC · Direct downloads · Manifest

sast is a tiny launcher. Installing it is instant; the first time you run it, it downloads a self-contained SAST engine binary that matches your operating system, verifies its checksum, and caches it. Every run after that is native speed with no Python dependencies.

pip install sast
sast .                       # scan the current directory
sast ./src -f sarif -o out   # write a SARIF report into ./out
sast . --fail-on high        # exit non-zero on high+ findings (CI gating)
sast --help                  # full engine options

▶ Watch the overview video  ·  Read the SDLC walkthrough →

Coverage at a glance: 1,750+ FP-validated rules · native AST + cross-file taint on 16 languages (regex for 40+) · 230+ secret rule packs with live key validation · SCA across 8+ ecosystems + container/OS packages · ~24,000 CMS advisories · web-shell & malware signatures · IaC (Terraform/K8s/Docker/ CloudFormation) · SARIF / JSON / HTML.

Benchmark (reference suite): ≈ 98.8% SAST coverage / 96.7% recall at a 0.0% false-positive gate, verified against known-vulnerability labs across all supported languages. Engine build 2026.10.10 (v3.4.9).

Supports Linux, macOS and Windows (x86-64). On Apple Silicon the macOS binary runs under Rosetta.

Installing

Pick whichever fits your stack — every method lands the same engine.

Debian / Ubuntu — signed apt repo

curl -sSL https://insom.ai/apt/public-key.asc | sudo gpg --yes --dearmor \
  -o /usr/share/keyrings/insomnia-archive-keyring.gpg
echo "deb [signed-by=/usr/share/keyrings/insomnia-archive-keyring.gpg] https://insom.ai/apt stable main" \
  | sudo tee /etc/apt/sources.list.d/insomnia.list
sudo apt-get update && sudo apt-get install -y sast

Direct download from insom.ai (no package manager)

A single self-contained binary — handy for air-gapped boxes and minimal CI images:

curl -sSL https://insom.ai/latest/sast/linux  -o /usr/local/bin/sast   # Linux x86-64
# curl -sSL https://insom.ai/latest/sast/macos -o /usr/local/bin/sast   # macOS
chmod +x /usr/local/bin/sast
sast --version

Windows: download https://insom.ai/latest/sast/windows, or grab any build from the downloads page. Files + SHA-256 are listed in the manifest.

Notes on the pip install

pip install sast creates a sast command (Linux/macOS: <prefix>/bin/sast, Windows: <prefix>\Scripts\sast.exe). For the command to be found, that directory must be on your PATH. Inside a virtual environment:

python -m venv .venv
# Linux/macOS:
source .venv/bin/activate
# Windows:
.venv\Scripts\activate
pip install sast

If sast is "not recognized" / "command not found" after a pip install --user, the per-user scripts dir isn't on your PATH. Either add it, or just run it as a module — this always works regardless of PATH:

python -m sast .

• Windows per-user scripts dir: %APPDATA%\Python\Python3XX\Scripts
• Linux/macOS per-user scripts dir: ~/.local/bin

Run with Docker

A prebuilt, multi-arch (amd64 + arm64) image ships the engine baked in — nothing to install, ideal for CI/CD. Available on both registries:

• Docker Hub: dominators/sast
• GHCR: ghcr.io/vulnz/sast

docker pull dominators/sast:latest        # or: ghcr.io/vulnz/sast:latest

Scan local source code — report saved to your folder. Mount your project at /src; the reports land in ./reports on your machine (not inside the container, which --rm discards):

docker run --rm -v "$PWD:/src" dominators/sast:latest /src -f html,json,sarif -o /src/reports --fail-on high

# Windows PowerShell
docker run --rm -v "${PWD}:/src" dominators/sast:latest /src -f html,json,sarif -o /src/reports --fail-on high

-o /src/reports writes to the mounted directory, so ./reports/ on the host holds the HTML/JSON/SARIF output. --fail-on high exits non-zero on high+ findings (CI gating). Drop -f/-o for a quick terminal-only scan.

Scan a remote repository — pass a URL and the engine shallow-clones, scans, and cleans up (no git clone, no volume mount needed):

docker run --rm dominators/sast:latest https://github.com/owner/repo --fail-on high
# private repo:
docker run --rm -e GITHUB_TOKEN=ghp_xxx dominators/sast:latest https://github.com/owner/private-repo

In GitHub Actions the image works as a container step:

- name: SAST
  run: docker run --rm -v "$PWD:/src" dominators/sast:latest /src -f sarif -o /src/out --fail-on high
- uses: github/codeql-action/upload-sarif@v3
  with: { sarif_file: out }

Tip: pin a version (ghcr.io/vulnz/sast:1.8.2) for reproducible builds. The image runs as a non-root user and needs no API key, account, or telemetry.

What it scans

A single self-contained binary — no Python dependencies and no external tools shelled out (no semgrep / trivy / bandit; everything runs in-engine):

• Code vulnerabilities — a native tree-sitter AST + taint engine across 16 languages (Python, JavaScript, TypeScript, Java, Go, Ruby, PHP, C, C++, C#, Rust, Kotlin, Scala, Swift, Lua, Shell) plus regex coverage for 40+, with intra- and cross-file taint that follows include / require / import relationships (source in one file → sink in another).
• Secrets & API keys — entropy + 230+ vendor rule packs, with optional live validation that labels each secret validated / invalid / not validated against its provider.
• Vulnerable dependencies (SCA) — OSV / advisory matching across npm, pip, Maven, Go, Composer, Cargo and more (covers npm audit-style checks).
• Vulnerable JS libraries — RetireJS-style detection of bundled jQuery, AngularJS, lodash, Bootstrap, Handlebars, … versions with known CVEs.
• CMS vulnerable components — WordPress / Joomla / Drupal / Magento plugins, themes and core matched against ~24,000 advisories.
• Web shells & malware — PHP / ASP(X) / JSP / shellcode signatures (c99, r57, WSO, China Chopper, …), obfuscated payloads, reverse shells, and suspicious double-extension uploads (shell.php.jpg).
• IaC / cloud misconfiguration — Terraform, Kubernetes, Docker, CloudFormation, …
• Output as HTML, TXT, JSON or SARIF 2.1.0 (drops straight into GitHub code scanning), with CI exit-code gating via --fail-on.

How it compares

Most free scanners do exactly one thing — Bandit is Python-only SAST, Checkov is IaC-only, Gitleaks is secrets-only, Trivy is SCA/containers. The all-in-one platforms (Semgrep, Snyk) put taint analysis, secrets and SCA behind a paid tier. sast does the whole lot in a single free binary — no external tools shelled out, runs offline.

✅ built-in · ◑ partial / paid tier · ✗ not offered

Reflects each tool's free/open-source offering as of mid-2026; paid platform tiers may add capabilities. In short: instead of stitching together Bandit + Gitleaks + Trivy + Checkov (and paying Semgrep/Snyk for taint + SCA), you get one free engine that covers all of it.

Editor & CI integrations

Same engine, everywhere:

• VS Code — the Insomnia SAST extension: inline hints as you type, a Vulnerabilities panel, an All Issues dashboard, mark-as-false-positive, and a git pre-push gate. → VS Code Marketplace · Open VSX (VSCodium / Cursor / Windsurf / Gitpod)
• JetBrains (PyCharm / PhpStorm / WebStorm / GoLand / IntelliJ IDEA / RubyMine / CLion / Rider / DataGrip) — one plugin for all of them. → JetBrains Marketplace
• Visual Studio 2022 and more entry points on the plugin page.
• GitHub Action — drop SAST into any pipeline in one line, SARIF straight to the Security tab: SAST by Insomnia.
• CI (any) — pip install sast && sast . -f sarif -o out --fail-on high, then upload out/*.sarif to GitHub code scanning.

How it works

pip install sast lays down only a few KB of pure-Python launcher — no download happens at install time (that keeps offline/CI installs reliable). On first invocation the launcher:

1. Detects your OS → linux / macos / windows.
2. Fetches https://insom.ai/static/downloads/plugin_manifest.json and reads its sast.<os> entry (filename + sha256 + version).
3. Downloads the matching binary and verifies its sha256.
4. Caches it under your per-user cache directory and execs it.

Because the engine lives on the server, new engine releases reach users without republishing the pip package.

Staying on the latest engine

After the first download the cached binary is reused for speed. At most once per day (SAST_UPDATE_INTERVAL, default 86400 seconds) sast also asks insom.ai whether a newer engine is published and, if so, upgrades itself automatically. Update checks fail open — if you're offline or the server is unreachable, the cached binary keeps working. Set SAST_UPDATE_INTERVAL=0 to pin the cached version, or run sast self-update to force the latest at any time.

Launcher commands

Environment variables

Default cache locations:

• Linux: ~/.cache/sast/bin
• macOS: ~/Library/Application Support/sast/bin
• Windows: %LOCALAPPDATA%\sast\bin

Use in CI (GitHub Actions)

- run: pip install sast
- run: sast . -f sarif -o sarif-out --fail-on high
- uses: github/codeql-action/upload-sarif@v3
  with:
    sarif_file: sarif-out      # the action accepts a directory of .sarif files

sast exits non-zero when findings meet your --fail-on threshold, failing the build.

Server-side manifest format

By default the launcher reads insom.ai's plugin_manifest.json, whose sast section lists the latest per-OS build:

{
  "sast": {
    "windows": { "filename": "insomnia-sast-windows-x64.exe", "sha256": "<hex>", "version": "1.0.0", "uploaded": "<iso8601>" },
    "linux":   { "filename": "insomnia-sast-linux-x64",       "sha256": "<hex>", "version": "1.0.0", "uploaded": "<iso8601>" },
    "macos":   null
  }
}

filename is resolved relative to the manifest URL. The launcher also accepts a simpler { "version", "platforms": { "<os>": { "url", "sha256" } } } shape if you self-host via SAST_MANIFEST_URL. sha256 is enforced when present; a null OS entry means that build isn't published yet.

© CQR Cybersecurity LLC. The sast launcher is open source; the SAST engine binary it downloads is proprietary. See https://insom.ai.