Repository navigation
v0.2.0 — Live OSV vulnerability data + major false-positive cleanup
Argus is an open-source AI Security Engineer — it maps your project, finds
vulnerabilities, explains each one like a senior security engineer, and can fix
them. Runs fully offline and free; your code never leaves your machine.
What's new in 0.2.0
- Live vulnerability data (OSV). The dependency scanner now checks your
packages against the public OSV database — thousands of real,
current CVEs instead of a small bundled list. Only package names and versions
are sent (never your source), and it falls back to offline mode automatically. - Major false-positive cleanup. Validated against real-world projects and
removed the noise that erodes trust:- Ignores hashes explicitly marked
usedforsecurity=False. - Ignores safe
pickle.loads(pickle.dumps(...))round-trips. - Downgrades findings in test/fixture code instead of screaming Critical.
- Correctly treats test certificates/keys as low-risk.
- Result on
psf/requests: 18 findings (4 Critical, 11 High) → 8 (all Low/Medium, all test code).
- Ignores hashes explicitly marked
- De-duplicated CVEs so one vulnerability is reported once, at its highest severity.
What Argus does
- Scanners: secrets, dependencies (OSV), static code analysis (SAST), and
infrastructure (Docker/Kubernetes/Terraform). - Explains every finding — why it's exploitable, attacker's view, business
impact, CWE/OWASP mapping, and remediation. - Attack Simulation Mode — safe, non-executing walkthroughs of how each flaw
would be exploited and how the fix blocks it. argus fix— applies verified fixes on a branch and opens a pull request.- Reports: JSON, SARIF (GitHub Code Scanning), Markdown, HTML, CSV.
- Bring your own model — offline by default; optional Anthropic, OpenAI, or
local Ollama.
Install
pip install argus-appsec