Virgil is an AI-assisted security audit platform for source code. Security at this moment essential for vibe coded applications and indie developers. It runs the scan in a sandbox, clusters duplicate findings, drops unreachable dependency CVEs, and ranks what to fix this week, with explanations grounded in your actual code. Open source, pay per audit, 3 free scans. You can find the web app at virgilhq.app, or download the cli: pipx install virgilhq&&virgil login&&virgil scan
No reviews yetBe the first to leave a review for Virgil
Maker
📌
Hey PH — maker here.
I built Virgil because every security scanner I'd used had the same failure mode: it finds everything, so you stop reading it. 200+ findings on a codebase, most of them either duplicates of the same root cause or CVEs in packages your code never actually imports.
What Virgil does differently:
- Locked-down sandbox (--network=none, read-only, non-root)
- Clusters findings that share a root cause into one item instead of N
- Checks whether a vulnerable dependency is actually reachable from your code (5 language families). Unreachable CVEs get demoted, not deleted
- Ranks what survives into a short queue with a one-line rationale each, using CISA's known-exploited-vulnerabilities list to outrank theoretical criticals over real-world-exploited ones
- Grounds every explanation in the actual code slice, so it's not just generic advice
- You can talk to Virgil, ask it for advice, what the findings mean.
3 free audits on signup -> no card required, then pay per audit. Nothing recurring.
The repo's open source (Apache-2.0) if you want to see how the pipeline works: https://github.com/ayaanmaliksgi.... Would love feedback, especially from anyone who's fought with false-positive fatigue on SAST/SCA tools before.
A bit more on how it works: it runs the scan in a locked-down sandbox (--network=none, read-only, non-root), then does the part most scanners skip, it clusters findings that share a root cause into one item instead of listing every call site separately, and checks whether a flagged dependency CVE is actually reachable from your code before it earns a spot in the queue. Unreachable ones get demoted, not deleted, so nothing's hidden, it's just not blocking your week. Your raw secrets never reach the LLM -> code context does, redacted first. Your privacy and security is important to us, and we do not break it.
Everything you see explained is grounded in the actual code slice too, the LLM isn't freelancing, it's citing what's really in your repo, and there's a validator that rejects anything that looks like a payload or exploit walkthrough.
3 more free audits have been credited to your account, if you want to point it at another repo. What repo did you run it against, out of curiosity?
1. You give it a link to your repo (or upload a zip) on either the web app or the command line interface. 2. It runs a few well-known security scanners on your code, safely, inside a locked-down sandbox. 3. Instead of dumping every single warning on you, it groups duplicates together and ignores the ones that don't actually apply to your code (like a vulnerability in a package you never even use). 4. You get a short list back: "fix these this week," with a plain-English reason for each one.
Takes about 3 minutes. Happy to answer anything more specific!
@ayanm I've tried it and it was very helpful
Tell me more about it!!!!!!!
NOWWWWW
@huzaifa_anwar Thanks, Huzaifa, glad it was useful!
A bit more on how it works: it runs the scan in a locked-down sandbox (--network=none, read-only, non-root), then does the part most scanners skip, it clusters findings that share a root cause into one item instead of listing every call site separately, and checks whether a flagged dependency CVE is actually reachable from your code before it earns a spot in the queue. Unreachable ones get demoted, not deleted, so nothing's hidden, it's just not blocking your week. Your raw secrets never reach the LLM -> code context does, redacted first. Your privacy and security is important to us, and we do not break it.
Everything you see explained is grounded in the actual code slice too, the LLM isn't freelancing, it's citing what's really in your repo, and there's a validator that rejects anything that looks like a payload or exploit walkthrough.
3 more free audits have been credited to your account, if you want to point it at another repo. What repo did you run it against, out of curiosity?
Actually I don't understand how it works
Some_one help me.
@abdullah_randhawa2 Sure, simple version:
1. You give it a link to your repo (or upload a zip) on either the web app or the command line interface.
2. It runs a few well-known security scanners on your code, safely, inside a locked-down sandbox.
3. Instead of dumping every single warning on you, it groups duplicates together and ignores the ones that don't actually apply to your code (like a vulnerability in a package you never even use).
4. You get a short list back: "fix these this week," with a plain-English reason for each one.
Takes about 3 minutes. Happy to answer anything more specific!