GitHub pull request security check · impacted tests

Merge with evidence, not hope.

AI writes code fast, and a change can open a door an attacker may find or expose your customers' personal data. CodeMD checks each pull request for new risky paths and runs the tests connected to the change, so you can fix problems while they're cheap.

Call-Path Security, Impacted Tests for your existing tests, and local test creation are free. Pro adds AI test commands, AI-written tests and higher limits. Your tests run in your own GitHub Actions.

Already installed? Open your checks →

An illustration of what appears in a pull request's checks list on GitHub.

Then see it all in one dashboard.

Every check across every pull request rolls up into your My checks dashboard: what's open, what's covered, and which endpoints are exposed.

codemd.dev/prchecks
acme/shopupdated just now · last 30 days
12pushes checked (last 24 h)
92%passed both checks
1new security issue today
0code bugs open
96%tests passed today (118 run)
88%changed lines tests ran (target 80%)

Needs attention: unsafe or high-impact security issues

FunctionImpactAI reviewWhyFound
export_orders
orders/export.py:42
highunsafePOST /orders/export reaches a shell command2 h ago
render_invoice
billing/pdf.py:88
mediumunsurecustomer-supplied name reaches a file write1 d ago

An illustration of the dashboard. Names and numbers are examples. Open your own →

Fewer security surprises. Fewer bugs reaching customers.

CodeMD looks for new ways an outsider could reach risky code or personal data, and runs the tests connected to the change. Each check below says what it looks for and how that can help you.

Free

CodeMD Call-Path Security

A change can quietly give an outsider a route to something dangerous in your code. This check looks at what this change added and fails the pull request when it finds a new route, so you can close it before it ships. It starts from what an outsider can send: web requests, WebSockets, messages from AI agents, and similar inputs. What it looks for, and how that helps:

  • Commands run on your server: helps lower the chance of an attacker running their own commands, a common route to stolen data or a taken-over server
  • SQL built from text (injection): helps lower the chance of customer records being read or changed through a form field or web address
  • eval and unsafe deserialization: helps keep an attacker's input from being run as code inside your app
  • HTML written into a page (XSS): helps keep one visitor's script from running in another visitor's browser, where it can steal a session
  • Files served from a variable path (path traversal): helps lower the chance of private files, such as config and keys, being read by someone who shouldn't
  • Open redirects: helps keep your own links from sending customers to a phishing site
  • File writes and deletes, git operations, outbound requests: helps flag outside input that could overwrite files, drive changes to your repository, or make your server contact places it shouldn't
  • Personal data: flags credentials, government IDs, health, payment and bank details, biometrics, identity, contact and location data that a function returns, logs or puts in a URL where outside input can reach it, and notes whether a sign-in check is on the path. Helps you catch a possible leak of customer data before it becomes a breach notice, a fine or lost trust
  • Web endpoints: lists new or wider endpoints and whether each shows a sign-in check, so you can see which doors are open to the public. This doesn't fail the check
  • Shows the whole path from the outside input to the risky call, not just the changed line, so a reviewer can judge it quickly
  • Only new problems fail the check, so review time goes to what you're about to ship, not to a backlog
  • Covers Python, JavaScript/TypeScript, Java, Kotlin, Go, C#, Rust, PHP, Ruby and C/C++. The check itself uses no AI
VS Code · Free

Call-path test creation

When a changed or risky path has no focused test, the free VS Code plugin creates one and runs it in an isolated copy of your workspace before you commit.

  • Creates call-path, contract, multi-caller, replay and new-function tests
  • Targets high-impact paths instead of guessing from filenames
  • Lets you review and commit the generated test
  • Committed tests join your existing suite and can run on every PR
GitHub · Free + Pro

CodeMD Impacted Tests

A change can break something several calls away from the line you edited. This check finds the tests connected to your change and runs them, so problems are more likely to surface before your customers see them.

  • Runs the tests that matter for this change, found by following call paths. That can mean less waiting on CI and fewer GitHub Actions minutes than running the whole suite
  • Checks the tests actually ran the changed lines (Python, JavaScript/TypeScript, Go, Rust, PHP and Ruby), so a green result says more than "something passed"
  • Lists each changed function with its test, the result and how many changed lines the test ran, so a reviewer can see the evidence without running anything
  • Runs in your own GitHub Actions. Your code doesn't run on our servers
  • Includes tests you created and committed, so the tests you add keep paying off on later changes
  • Runs existing tests for Python, JavaScript/TypeScript, Go, Rust, Java, Kotlin, C#, PHP and Ruby
  • Free for existing tests. Pro adds AI-found test commands and AI-written tests
  • Waits for the security scan, then runs, so results land together

Working in three minutes.

No config file, no new workflow to maintain.

1

Install the app

Click install, pick the repositories to check. Access is read-only for code.

2

Open or push a PR

Both checks appear in the pull request's checks list within minutes.

3

Fix it or mark it safe

Click "View more details", sign in with GitHub, and mark a checked path "Verified safe" with a reason. Or add a codemd-ignore comment in the code, so a reviewer sees it in the diff.

Why install it.

Helps you keep up with AI-written code

Coding agents change a lot of code, fast, and a reviewer can miss what a change makes reachable. The checks look at that for you, which can lower the chance of a security hole or a leak of customer data reaching production.

Saves review time and rework

The risky call is often several functions away from the line that changed. CodeMD follows the call path so the reviewer doesn't have to, and a problem fixed before merge usually costs far less than one fixed after a customer or an attacker finds it.

Evidence you can act on

Each finding shows the actual path through your code, so you can decide quickly whether to fix it or mark it safe. A test result shows whether the changed lines ran.

In our own test, the free check flagged 12 of 14 published command-injection bugs in open-source MCP servers. See the results, including the 2 it missed →

Free to start.

Call-Path Security, Impacted Tests for your existing tests and VS Code test creation are free. Pro adds AI test commands, AI-written tests, higher limits and the AI security review.

Free

$0 forever
  • CodeMD Call-Path Security on every push and PR
  • CodeMD Impacted Tests runs your existing tests, up to 10 checks an hour per installation, behind Pro in the queue
  • VS Code call-graph analysis and missing-test creation
  • Unlimited public repositories
  • Private repositories during early access
  • Free repository analysis

Install free

Pro

$4.99 per developer / month
  • Everything in Free
  • CodeMD Impacted Tests also finds test commands with AI and can write tests for files that have none, up to 20 test checks an hour
  • AI verdict in the check: real, safe or unsure, with a fix
  • First in the queue, 200 security checks an hour
  • 30-day free trial

See Pro

Questions.

Does CodeMD see or keep my source code?

The app reads your repository to analyse a commit; access is read-only. Tests run in your repository's own GitHub Actions, not on our servers.

Will it block my merges?

The checks appear in GitHub like any other. Whether a failing check blocks a merge is up to your branch protection rules.

What if it flags something that is actually safe?

Sign in on "My checks" and mark the path "Verified safe" with a reason, or add a codemd-ignore: the command is fixed comment (with your reason) inside the risky function or on the line above it. The comment shows in the pull request's diff, and the check lists what was dismissed. We review every mark to improve the scanner.

What are the limits on the free plan?

Impacted Tests runs your existing tests for up to 10 checks an hour per installation. Free checks queue behind Pro ones, and a test whose command only AI can work out is skipped with "finding its command needs AI". Pro raises the test limit to 20 an hour, finds those commands with AI and writes tests for files that have none. Security checks: up to 30 an hour per installation on Free, and 200 an hour per developer on Pro.

Which languages does it support?

Python, JavaScript/TypeScript, Go, Rust, Java, Kotlin, C#, PHP and Ruby for call paths, test discovery and running existing tests.

How do I remove it?

Uninstall the app from your GitHub settings at any time. The checks stop immediately.

See what your next pull request exposes.

Install CodeMD PR Checks free. Takes one click.

Install on GitHub