Questions, answered.
What Mallace finds, what it costs, where your code goes and what it won't do. Can't find your answer? Email [email protected].
What does Mallace do?+
Mallace scans GitHub repositories written in Python, JavaScript, TypeScript, PHP or Ruby for serious security bugs. For each suspected bug, it writes a regression test and runs it against your code. If the test fails, the bug is real, so Mallace writes a fix, checks that the test now passes, has a second model review it, and opens a pull request with the fix and the test.
Which vulnerabilities does it find?+
SQL injection, command injection and code injection (eval, plus exec in Python and instance_eval/class_eval in Ruby) in every supported language, and unsafe deserialization: yaml.load and pickle in Python, unserialize in PHP, Marshal.load and unsafe YAML in Ruby, plus template injection in Go, cross-site scripting (reflected and stored) and path traversal in Python, JavaScript/TypeScript, PHP, Ruby and Go, and server-side request forgery in Python and JavaScript/TypeScript. That’s 40 rules today, each one proved with a test and fixed with a pull request. It also reports leaked secrets and vulnerable dependencies, and checks live websites. It doesn’t yet scan infrastructure or cloud configuration.
Does it check for leaked secrets and vulnerable dependencies?+
Yes, on every scan, free. It reports API keys and private keys committed to the repository (never printing the key), and package versions with known vulnerabilities from requirements*.txt, package-lock.json, composer.lock and Gemfile.lock, using the public OSV database. These are reported, not fixed: rotate leaked keys, and upgrade packages to the version shown.
Can Mallace attack my app like a hacker would?+
Yes, a sandboxed copy of it. mallace attack starts your web app from your pushed code in a Docker sandbox with no network, sends attack requests to that copy, and reports only attacks it proved with planted canaries: SQL, command, code and template injection, file reads outside the app, and cross-site scripting. It never touches your live site. With --fix it also fixes what it proved (injection, file reads and cross-site scripting), with a regression test that replays the attack, and opens a pull request for each verified fix. If the app needs Postgres, MySQL, Redis or MongoDB, Mallace starts a throwaway copy in the sandbox, and it can log in to test pages behind a login.
Can Mallace test my live website?+
Yes, with read-only checks. Mallace looks for HTTPS and certificate problems, missing security headers, insecure cookies, exposed files like .env and .git, and outdated WordPress, after you prove you own the site. It works on any site, including Wix, Squarespace and Shopify, without GitHub. Run it from the app (Free covers 1 website, Pro 10) or with mallace site check. It doesn’t attack the site, so it isn’t a penetration test.
Which languages are supported?+
Python, JavaScript, TypeScript, PHP, Ruby and Go, including Ruby on Rails apps and WordPress plugins. Java findings are proved and fixed using the JDK alone. Rust, C# and other languages aren’t covered.
Does Mallace merge anything automatically?+
No. Every fix arrives as a pull request for you to review and merge yourself.
Is scanning really free?+
Yes. Scans are unlimited on every plan and don’t need a license key. Only proving and fixing bugs uses your plan’s AI allowance.
How much does it cost?+
Free includes unlimited scans and fixes on one repository, about 3 a month. Pro is $29 a month or $290 a year for unlimited repositories and about 50 fixes a month. The price is per account, not per developer.
What counts as a "fix" in the allowance?+
The allowance is AI spend, not a fixed count: $1 a month on Free and $15 on Pro. A typical fix uses about $0.30 of it, which is where "about 3" and "about 50" come from. Findings that turn out not to be real usually cost less.
Does my code leave my infrastructure?+
Scans and tests run on your GitHub Actions runner or your machine. To write a test, fix or review, Mallace sends the code around that one finding (with common credential formats redacted) to its service, which passes it to the AI model for that call and doesn’t store it.
Is my code used to train AI models?+
No. Mallace doesn’t train models on your code, and the AI provider processes it under its API terms.
Do I need my own OpenAI key?+
No. Fixes run on the Mallace service with your Mallace license key. Older workflows that passed an OpenAI key get an update pull request from the app.
What happens if the AI writes a bad fix?+
A fix only becomes a pull request if the proving test passes, the code compiles and an independent reviewer model approves it. If not, Mallace tries again or reports that no fix passed. You still review every pull request before merging.
What if a finding is a false positive?+
If the regression test passes on your current code, Mallace reports the finding as "not reproduced" and changes nothing.
Does it work with GitLab or Bitbucket?+
Not yet. Mallace works with GitHub repositories only: the app, the GitHub Action and the CLI all open issues and pull requests on GitHub.
Can I run it locally?+
Yes. pip install --pre mallace, then mallace fix --scan-only for a free scan, or mallace activate and mallace fix to get fixes. It needs Docker and the GitHub CLI.
How do I cancel?+
Click Manage billing in the app. You keep Pro until the end of the period you paid for, and you aren’t charged again. Charges can be refunded within 14 days; see the refund policy.
Will it open duplicate pull requests?+
No. A finding that already has an open Mallace pull request is skipped, even after new commits.
Who makes Mallace?+
Mallace is made by CodeFront Labs. Contact [email protected].