Privacy
Your code is not ours, and most of it we never hold at all.
MLReps is a digital product. A data request, a deletion request or a security report reaches us at connect@mlreps.com — the same address published in our security.txt.
What is true, and why it is true
- Your source code is never stored.
- No column exists that can hold source, a diff, a patch or a token — retaining one would take a database migration, and a build gate fails any log call that could receive source. On the command-line path your code never leaves your machine at all. On the GitHub App path it is unpacked onto an analysis worker, read inside a sandbox with no network, and deleted when that one analysis ends.
- Nothing about your code is sent to an AI model — not your source, and not metadata about it.
- No file path, no line number, no operation name, no count and no finding of ours leaves for a model, and no model writes a word you read. Three things make that so rather than one: the one screen that sent a model anything no longer does, no model key is configured in the service at all, and a build gate counts every place in the code that could reach one. If we ever turn on the assistant that would send a question and its evidence to be worded, /security says what that sends and this page will say it before it happens.
- Nothing in your browser tracks you.
- No analytics script loads on any page of this site — there is no measurement id configured in any build — no cookie is set to measure you, and no third-party tag runs anywhere, inside the product or outside it. Counting how many people visit is our own, and it is four things: the page path, the hour, which kind of device, and which site you arrived from. The first three come from the request your browser makes anyway. The last one your browser tells us, because only it knows — and we keep it as one of ten words such as “search” or “direct”, never the address. No cookie, no identifier of you, nothing that links one visit to the next, and nothing stored about where you went inside the product.
- We write to your repository in one place only, and only when you ask.
- One instrumentation file, on a branch of its own, as a pull request you review and merge — opened only when you ask for it in MLReps. Nothing else: no push to your default branch, no merge, no second file, no second pull request. A fix is never committed; it stays a patch you apply yourself. The analysis worker that reads your code asks for read access and never holds the write, the commit is made from our web host with a token that names one repository and is revoked before the call returns, and our pull-request check can only report neutral, so it is incapable of blocking your merge.
- Nobody is ever ranked.
- There is no per-developer score, no leaderboard, and no screen that names who merged what — permanently, by decision. A record readable by whoever writes the performance review stops being true, so the endpoint does not exist rather than being access-controlled.
- One organisation can never read another's data.
- Every query against a table that belongs to a tenant is scoped to that tenant, and a build gate checks every one of them — not a review habit, a check that fails the build.
- Credentials we find in your code are removed before anything leaves the process.
- A detector suite runs over anything on its way out, including a key split across a line break, and replaces it whole, so no fragment leaves and no folded line is shown back to you.
- In transit and at rest, it is encrypted.
- TLS 1.3 everywhere, including between the worker and the control plane; encrypted storage, with per-tenant key separation for the knowledge plane.
What we keep, and for how long
- Repository names
- The full name only — "acme/payments" — for the repositories you selected when you installed the App.
- Kept: Until you uninstall, which removes them at once.
- Findings and the architecture map
- What each rule found, at what severity, and the shapes the report carries: file paths, URL patterns, table, package and operation names, and line numbers. No source text. This is the one thing we genuinely hold at rest, and saying so is the point of this row.
- Kept: Until you ask us to delete it. Uninstalling the App does not erase it — see below.
- Your account
- An email address and a password hash, and which organisation you belong to.
- Kept: Until the account is deleted.
- A payment reference, if you pay
- The reference the provider gave your payment, the amount in whole cents, and when it was credited — because that reference is what stops one payment being credited twice. Never a card number, never the last four digits, never a billing address: those go from you to the provider and never through us.
- Kept: For as long as the balance exists. The ledger is append-only, so a credit is never edited or removed — a correction is another entry beside it.
Who else receives anything
Three, and no others. Each one is named because a page claiming it shares nothing with anybody would not be true, and naming them is what should make the rest of this page believable.
- GitHub
- The App reads your repositories under the scopes GitHub shows you on its own consent screen. It is where your code already is.
- Amazon Web Services
- Hosting for the control plane and the analysis worker.
- Google Analytics
- Not running on this site today — there is no measurement id configured, in any build, so nothing is sent and no script loads. If we turn it on, this is exactly what it would be: on the public marketing pages only — the ten pages in our sitemap — and never inside the product. It never runs on any console or admin screen, on the page that lists your repositories, on the sign-in or password reset page, on the page a terminal sends you to, or on the page that sets up your first scan. What it never receives, on the pages where it would run: the query string of the address, the page you came from if that page was one of ours, any repository or organisation name, any account identifier, and any email address. What it would receive is the page's own path, where you arrived from if that was another site, and an approximate region. Consent is set to denied before anything loads and there is no banner to grant it, so no analytics cookie is stored on your device and nothing links one visit to the next.
A payment provider is not on this list, because we never send it anything about you. Paying happens on the provider's own page: the button sends your browser there, your card details go from you to them, and they never pass through us. What reaches us afterwards is the payment reference, typed in by somebody here, which is what turns a payment into a balance. There is no card number to leak here because there is no payment integration at all.
What is not done yet
A page that states only the reassuring half is the thing this product exists in opposition to. These are true as well.
- Uninstalling does not erase your history.
- It removes the connection and your repository names at once, which withdraws our access. Your findings and architecture map stay, deliberately, so that reinstalling next week does not find your record gone. If you want them gone, ask us and we will delete them.
- There is no self-serve delete, and no self-serve export.
- Both are done by hand on request today. We would rather say that than describe a button that is not there.
- Operator access is not yet mechanically blocked.
- Our own staff are held to aggregates by how the admin screens are built, and every one is restricted to counts and buckets that cannot name a customer. The stronger control — where reading a specific customer's data requires a time-boxed grant from that customer and is logged where you can see it — is specified and not yet built. Until it is, this is a practice, not a mechanism, and it is listed here instead of being left out.
Asking us to delete or send you your data
Write to connect@mlreps.com and we will delete everything we hold on your organisation, or send it to you. Both are done by hand today; neither is a button yet, and this page will say so until it is one.