Privacy
What CodeMD does with your code and data on codemd.dev. Last updated November 14, 2026.
In short
- We use your data only to run CodeMD for you. We don't sell your data, and we don't use your code to train AI models.
- We collect as little as we can, and we delete it when we no longer need it or when you ask.
- Email contact@codemd.dev to see, correct or delete your data.
Your account
- You sign in with GitHub. We keep your GitHub user ID, username, and the name, profile picture and email address GitHub shares with us, to show who's signed in and to link your subscription to you.
- We keep the repositories you've given CodeMD access to, so we can show and analyze them for you.
- Sign-in uses cookies on codemd.dev: one keeps you signed in, and short-lived ones protect the GitHub sign-in. We store sign-in sessions only in a scrambled (hashed) form.
- If you send us a message from codemd.dev, we keep your message and the contact details you give us, to reply to you.
- If you join the CodeMD Pro waitlist, we keep the email address or phone number you provide so we can tell you when CodeMD Pro launches. We don't store your IP address with the signup, share the list with third parties or use it for unrelated marketing. You can leave the waitlist at any time by emailing us.
Payments
Stripe handles payments for CodeMD Pro. Your card details go to Stripe, never to us. We keep your Stripe customer ID and your subscription's status, plan and renewal date.
Repository analysis on codemd.dev
When you analyze a repository on codemd.dev, CodeMD downloads its code to build the analysis. The downloaded copy is working data on our servers and is not kept as a repository copy: the downloaded ZIP is deleted as soon as it is unpacked, and the unpacked files are deleted after 7 days without use. We keep what the analysis produces so you can come back to it, and that includes some text taken from your code:
- Code structure: function and class names, file paths, line numbers, the call graph between functions, and numeric "embeddings" that let search find related functions. Function bodies are not stored in these files.
- Text from your repository: code comments and TODO notes, text and documentation files (such as README and config files) and the text shown in web pages. Long files are cut off, so only part of a file may be kept. Files made to hold credentials (
.env, key files,secretsandcredentialsfiles, Springapplication-*.properties) are never read, and key-, token- and password-shaped values in what is kept are masked. - Results: feature lists, security findings, test results and the bugs and defects found.
Comments and documentation can contain anything you wrote in them, so don't put secrets there. If you run the analysis while signed in with GitHub, the results can be seen only by people who can see that repository on GitHub. Results for a public repository analyzed without signing in are kept in a separate private store, and are opened only through short-lived links the site creates. AI features, when you use them, send the code they need to an AI provider, as described under "Services we use" above.
You can delete a repository's analysis any time. With push access to the repository, press Delete analysis on your dashboard; otherwise email contact@codemd.dev and we'll delete it.
Services we use
These services process data for us, only to run CodeMD:
- GitHub: sign-in, and access to the repositories you choose.
- Render: hosting codemd.dev and running analyses and checks.
- Supabase: our database.
- Stripe: payments.
- OpenAI and Anthropic: Pro's AI features, which receive only the code those features need. Their APIs don't use this data to train their models.
- Mailjet: sending the messages you send us from codemd.dev.
We may also share data if the law requires it, to protect CodeMD or its users from harm or abuse, or as part of a merger, acquisition or sale of CodeMD, where these promises continue to apply.
CodeMD PR Checks
The single CodeMD PR Checks GitHub App posts CodeMD Call-Path Security and CodeMD Impacted Tests on pull requests and supported pushes in the repositories you choose.
What the app can access
- Read your repository's code, pull requests and metadata.
- Write check results. For test execution, it can also write only to the dedicated
codemd-test-runsbranch, as described under CodeMD Impacted Tests below. It doesn't comment or change repository settings.
What happens to your code
- For each commit it checks, CodeMD downloads that commit's code to its servers, builds the call graph, and looks for paths from outside input to risky calls.
- The downloaded code is deleted as soon as the scan finishes. Scan folders a crash leaves behind are swept up within hours.
- CodeMD keeps a compact call graph (function names, file paths and file fingerprints, not code) so the next commit can be checked by comparison instead of a full rescan.
- The free check runs no AI: your code is not sent to any AI model.
- With CodeMD Pro, when the person who pushed has a paid subscription and the change adds a new risky path, only the files on that path (at most 8) are downloaded, sent to OpenAI for a verdict, and deleted. The verdict is added to the check. OpenAI's API doesn't use this data to train its models.
What CodeMD keeps
- The list of risky functions for each checked commit: function names, file paths, line numbers, the kind of risky call, and the call path from the input. This is so the next commit can be compared with it. Up to the 50 most recent commits per repository are kept, and older ones are deleted.
- Code-weakness findings (such as a hardcoded key or debug mode left on): the file, line number and a short line of code. Keys, tokens and passwords in that line are masked before it is saved.
- A queue entry for each check: repository owner and name, installation ID, commit IDs, the GitHub account ID of who pushed, and whether the check finished.
- The reason in a
codemd-ignorecomment, when your code dismisses a path: it's kept with that commit's list and shown in the check. - Server logs name the repository and commit when a check fails.
Who sees the result
The check result is posted to GitHub, so anyone who can see the repository on GitHub can see it, including the function names and file paths it lists.
Stopping it
Uninstall the app, or remove a repository from it, in your GitHub settings. CodeMD then can't read that repository and stops checking it, and GitHub tells CodeMD, which deletes that repository's kept lists of risky functions and queue entries right away.
Limits
The check is free and shared, so each installation gets a limited number of checks per hour, large repositories aren't scanned, and checks may be skipped when too many are waiting. A skipped commit gets a neutral "not scanned" result instead of a pass or fail.
CodeMD Impacted Tests
The CodeMD PR Checks GitHub App posts CodeMD Impacted Tests on pull requests and pushes to the default branch. Free runs existing tests whose commands CodeMD can determine without AI. Pro can use AI to determine additional commands and create a test when a changed file has none.
What the app can access
- Read your repository's code, pull requests and metadata, and its GitHub Actions runs.
- Write check results and the
codemd-test-runsbranch with its workflow file. CodeMD requires repository write permissions because GitHub doesn't support branch-scoped GitHub App permissions. The app only writes to the dedicatedcodemd-test-runsbranch and never modifies your source code, pull-request branches or default branch.
Optional: enforce the branch limit in GitHub
A repository administrator can make GitHub reject any attempt by CodeMD to update another branch:
- Open the repository on GitHub, then go to Settings → Rules → Rulesets → New ruleset → New branch ruleset.
- Name the ruleset (for example,
Protect branches from apps) and set its enforcement status to Active. - Under Target branches, include all branches, then add an exclusion for
codemd-test-runs. - Under Bypass list, add only the trusted people, teams or repository roles that must be able to update those branches. Do not add the CodeMD GitHub App.
- Enable Restrict updates. We also recommend Restrict deletions and blocking force pushes.
- Save the ruleset. Use GitHub's rule insights or test with a non-default branch to confirm normal developer and pull-request workflows still work before relying on it.
This ruleset protects the default branch and every pull-request branch it targets while leaving codemd-test-runs available to CodeMD. Ruleset availability depends on the repository visibility and GitHub plan. See the detailed CodeMD installation and branch-lockdown guide, GitHub's documentation for creating a repository ruleset, and the available rules.
What happens to your code
- For each commit it checks, CodeMD downloads that commit's code to its servers to find the existing tests for the files the change touched, then deletes it.
- When only AI can tell how a test is run, the repository's files are read by Claude or OpenAI to name the command (at most 3 times per check).
- This check runs your existing tests. It doesn't write tests: a changed file with no test is listed in the check as having none.
- Neither AI provider uses this data to train its models.
- The tests run in your repository's GitHub Actions, on the
codemd-test-runsbranch, so they use your Actions minutes. CodeMD reads the run's log to see which tests passed and whether they ran the changed lines.
What CodeMD keeps
- For each check: the repository, commit IDs, the changed file paths, the test file paths and commands, each test's result and whether it ran the changed lines, the Actions run link, the check's result, and the GitHub account ID of who pushed. It's deleted when you uninstall the app or remove the repository.
- The result in your repository's test history on codemd.dev (the Testboard), and a failing existing test in its Bugs, Issues & Defects list when the repository was analysed on codemd.dev.
Limits
Each installation can start a limited number of checks per hour, and each check runs a limited number of tests. A commit over the hourly limit gets a neutral "Hourly limit reached" result.
CodeMD Feature Impact
The CodeMD PR Checks GitHub App also posts CodeMD Feature Impact on pull requests. It is informational: it is always neutral and never fails a pull request.
What happens to your code
- CodeMD downloads the commit's code to its servers to build the call graph and work out which entry points the change reaches, then deletes it.
- To say which product features those entry points serve, CodeMD sends Anthropic's Claude names only: route paths, button or link text, and function names, together with your product & features list when you have one in the repository. No source code is sent. Anthropic's API doesn't use this data to train its models.
What CodeMD keeps
- The feature names and descriptions, and which entry-point names were labelled with which feature, so the next pull request doesn't ask again. It's deleted when you uninstall the app or remove the repository.
Website usage analytics
- codemd.dev records page views and successful actions (for example, an analysis finishing) in its own database.
- If you're signed in with GitHub, events are tied to your GitHub account ID. Otherwise they're anonymous.
- No cookies are set for analytics, and no IP addresses, search queries, prompts, code, error messages or query strings are stored.
- Only counts, yes/no values, fixed labels and repository names are recorded.
Keeping data safe
We use encrypted connections, limit who and what can reach stored data, keep access tokens out of our logs, and store them only encrypted or hashed. No service is perfectly secure, though, and we can't guarantee that data will never be exposed. If a breach affects your data, we'll tell you as the law requires.
Your rights
- You can ask to see, correct, export or delete your data, or object to how we use it. Email contact@codemd.dev and we'll respond within 30 days.
- Deleting your account also cancels Pro and deletes your analyses. We keep payment records as long as tax and accounting law requires.
- Depending on where you live (for example, the EU, the UK or California), the law may give you more rights. You can also complain to your data protection authority.
Children
CodeMD isn't for children under 13, and we don't knowingly collect their data. If you think a child has given us data, email us and we'll delete it.
Where data is processed
The services we use may process data in the United States and other countries, which may have different data protection laws from yours. Using CodeMD means your data may be processed there.
Changes
We may update this page. We'll change the date at the top, and for significant changes we'll give notice on codemd.dev.
Contact
Questions about your data, a request to delete it, or a security problem you found in CodeMD: email contact@codemd.dev or visit Support.