probus

Open-source AI vulnerability scanner powered by open models.

---

Probus started as an internal supply chain security scanning tool that proved itself extremely efficient by finding vulnerabilities in top open source packages (e.g. n8n, AI sdk, langraphjs and more). It is now open-source to help developers better secure their codebase & supply chain. Probus' edge lies in its ability to scale its scanning capabilities with open models (by using OpenRouter).

What it does

Probus harnesses 3 agents that:

• [Analyst] Analyze the codebase and pick key files for deep scanning (e.g. entry points, third-party surface, dangerous sinks).
• [Researcher] Scan each file, dig through its chains of calls, and write raw findings (potential vulnerabilities).
• [QA] Independently verify each finding, make sure it has a real attack vector, and write a report.

Quick start

npm install -g probus
probus

probus launches a local web server, opens your browser to it, and from there you:

• pick a repo to scan,
• enter / pick a model-provider API key,
• watch a live dashboard of the analyst → researcher → QA pipeline,
• and browse verified findings as polished markdown reports.

No CLI flags, no scan/view subcommands — everything happens on the page. The only options are launcher flags:

probus [--port <N>] [--no-open]

Flag	Default	What it does
--port	random	Pin the local server port (otherwise picks a free one)
--no-open	open	Skip auto-opening the browser; just print the URL

Model providers

Probus runs most (cost) effectively with open models using OpenRouter. It is still possible however to use other providers, such as OpenAI or Anthropic, albeit with higher costs.

You configure providers and keys directly in the web UI's Settings tab (or inline on the New scan screen). Keys are stored at ~/.probus/.env (chmod 600) and never leave the machine.

Defaults per provider

When you don't pass model overrides, these are the picks:

Provider	Primary default	Secondary default
openrouter	openrouter/qwen/qwen3.6-plus	openrouter/deepseek/deepseek-v4-pro
openai	openai/gpt-5.4-mini	openai/gpt-5.4
anthropic	anthropic/claude-sonnet-4-6	anthropic/claude-opus-4-7

Effort levels

Each scan picks an effort level that caps the number of files the analyst targets:

Effort	Files (approx)
low	50
medium	100
high	500

Cost

Probus splits work between two models so you only pay premium rates where it matters:

• Primary (~90% of tokens) — runs on every file. Pick something cheap and fast: qwen3.6, gpt-5.4-mini, sonnet-4.6.
• Secondary (~10% of tokens) — verifies findings. Pick something smarter: deepseek-v4-pro, gpt-5.4, opus-4.7.

Each file consumes roughly 1M input tokens. Approximate per-file cost by provider:

Provider	Cost / file	vs. open models
openrouter (open models)	~$0.50	1× (baseline)
openai	~$1.25	~2.5×
anthropic	~$5.00	~10×

Contributing

PRs welcome. See CONTRIBUTING.md for dev setup, scripts, and conventions.

Development

Local dev

git clone https://github.com/ItayRosen/Probus
cd probus
nvm use && npm install
npm run dev          # one command: Vite HMR for web/, tsx-watch reload for src/

npm run dev boots the server on :9091 in dev mode — backend changes hot-reload via tsx watch, frontend changes hot-reload via Vite middleware. No build step needed during iteration.

For a production-mode local run (serving the prebuilt bundle from dist/web):

npm run build && npm start

Architecture

┌────────────┐   files[]   ┌──────────────┐  findings[]  ┌───────────┐
│  Analyst   │────────────▶│   Primary    │─────────────▶│ Secondary │
│  (1 call)  │             │  (per file)  │              │ (per file)│
└────────────┘             └──────────────┘              └─────┬─────┘
                                                               │
                                                               ▼
                                                       reports/*.md

All three run as isolated query() sessions through the Claude Agent SDK, each with its own filesystem sandbox scoped to the repo being scanned.

Output layout

output/<repo-slug>/
├── analysis.json           # file list picked by the analyst
├── findings/
│   └── src__foo__bar.ts.json   # per-file findings (verified + unverified)
├── reports/
│   └── src__foo__bar.ts--1.md  # one Markdown report per verified finding
├── debug/
│   └── src__foo__bar.ts.log    # full agent transcript per file
└── processed-files.txt     # cache so reruns skip finished files

<repo-slug> is <basename>-<sha1(abspath)[:8]> so the same repo never collides with another.

License

Apache 2.0 — see LICENSE.