What we do with your code
Letting a tool read your source is a big ask. Here’s exactly what happens to your code - in plain English, no asterisks - and how to use your local coding agent if you prefer.
The short version
Six things to know
Read-only access
We can read your code. We can never write to it, commit, or open pull requests.
We never store your code
Your source is never saved to our database. Only the finished documentation is kept.
Deleted right after analysis
Each analysis runs in an isolated sandbox that’s destroyed the moment it finishes.
Never used to train AI
Your code is not used to train any AI model — not ours, not our providers’.
You own what we generate
The articles we produce are yours to edit, publish, export, or delete at any time.
You decide what’s public
Every article is a draft for your review. Nothing reaches your help centre without you.
Start to finish
Where your code actually goes
Four steps. Your code lives only in a temporary sandbox, never on a disk we keep.
Copy
When there’s something to document, your code is pulled into a fresh, isolated sandbox — just for that run.
Analyse
AI reads the code (or just the diff of what changed) and writes a draft help article. See “who sees it” below.
Delete
The sandbox is destroyed. The copy of your code goes with it. Nothing is retained.
Keep
Only the generated documentation remains, saved to your account for you to review.
Generate locally with WTFM
The steps above describe hosted repository analysis. You can also generate documentation from your local checkout with Codex or Claude Code, without granting SupportPages access to your repository provider.
Install the SupportPages command-line tool (WTFM), connect a help centre, then ask your coding agent to write an article. It uploads a draft for you to review and publish in SupportPages.
SupportPages receives article content, referenced images and a provenance manifest. Repository source files are excluded from uploads. Your coding agent uses its configured model provider during generation.
Set up WTFMFAQ
Straight answers
The questions security reviews actually ask, answered without hedging.
For day-to-day updates we only read the diff of what changed in a merged pull request. When you first point it at a repository (and for some richer outputs like walkthroughs) we take a temporary read of the wider codebase so the docs match your app. Either way it’s read-only, and it happens in the throwaway sandbox above. With WTFM, your coding agent reads the checkout using its configured model provider. SupportPages receives generated documentation assets.
No. Your source code is never written to our database — only the finished documentation is. One honest caveat: the diagnostic log of a generation run can contain short snippets of what the AI read. That log is attached to the specific draft to help debugging, and is replaced the next time that article is generated. We’re shortening its lifecycle further.
To generate your docs, your code is sent to our AI provider — Anthropic, Cursor, or OpenAI, depending on your project’s settings — and all are contractually bound not to store it after processing or use it to train models. Your repository is reached through GitHub. Everything we store lives in the EU (Finland). The full list is on our sub-processors page.
Never. Neither we nor our AI providers use your code to train models. It’s used only to generate your documentation, then it’s gone.
No. When you generate in the cloud we ask for read-only access through the GitHub App, and there is no code path in the product that can write to your repository. With WTFM, your coding agent uses its own permissions and model provider; SupportPages receives article content, referenced images and provenance.
We request the narrowest access we can, and the GitHub access we use is short-lived and scoped only to the repositories you connect. We store nothing of your source code, so there’s no repository of code to leak. Details are on our security page.
You stay in control of what goes public. By default, every article is a draft you approve before it’s ever published. If you turn on auto-publish, drafts go straight to your help centre — so if your code contains secrets or private data in comments or strings, we recommend keeping review on. Automatic secret-scanning is on our roadmap.
All stored data lives in the EU — Hetzner, Finland — with encrypted backups in the same region. Analysis is sent to our AI provider (Anthropic, Cursor, or OpenAI — all US-based) under Standard Contractual Clauses. See our DPA and sub-processors.
Yes. Disconnect a project or delete your account and we remove your projects, articles, and associated data. Copies may persist in encrypted backups for up to 30 days, then they’re gone too.
Honesty builds trust
What we don’t offer yet
We’d rather be straight with you than oversell. If your organisation needs any of the following today, talk to us before you commit - they’re in beta or on the roadmap, not generally available: