Things people usually want to know before they install. If yours isn't here, email us.
Do I need an API key? Does Foundry use my tokens?
Yes. Foundry is BYOK (bring your own key). You pay Anthropic directly for Scholar's analysis and proposals during setup. Once your team is generated, the agents run as part of your normal Claude Code usage. Foundry doesn't insert itself into runtime. Codebase scanning is local (Graphify, no tokens). MemPalace is local. The token cost is the wizard interview itself, typically $0.25 to $0.50 per setup.
What happens to my code? Does any of it leave my machine?
Your source files stay local. Graphify builds a semantic index entirely on your machine. Scholar reads summaries and references from that index, plus whatever you tell it in the interview. It never uploads your code wholesale. The only thing crossing the network is the structured context Scholar sends to the Claude API, and that data goes to Anthropic, not to us.
What if Scholar's proposals are wrong? Can I edit?
Yes, at every stage. The Review step shows every agent, hook, rule, and standard before anything writes. You can edit, remove, or reject any of them. After install, the Forge editor lets you keep editing, and every Foundry-managed item has a baseline restore-point so you can Reset to baseline if a later edit goes sideways. None of it is locked-in: it's all standard .claude/ config.
Is this lock-in? What if I uninstall?
No lock-in. Everything Foundry creates is standard Claude Code configuration: plain markdown agents, JSON hooks, and rules files. Uninstalling Foundry doesn't break your team; the config keeps working. Foundry tracks what it created via a manifest at .claude/.foundry-manifest.json so the uninstall flow knows exactly what to remove and what was user-authored.
Why "1.0 beta" and not just "1.0"?
The product is feature-complete and stable enough to call 1.0, but the paid-tier license-key system isn't ready yet, and we want to validate the launch flow with real users before charging anyone. Beta builds are time-gated to 30 days each, free during the beta period, with the next build available before the previous one expires. When license keys land, this becomes 1.0 stable.
Mac? Linux?
Windows-first. Linux is a stretch goal we hope to ship soon after launch (AppImage target). Mac is post-1.0. Apple code-signing and notarization add enough overhead that we're punting it until the Windows release proves out.
Is the "Windows protected your PC" warning safe?
It's safe. Foundry isn't code-signed yet (Authenticode certificates are on the roadmap), so Windows SmartScreen warns on first install. Click More info → Run anyway. Every released build is tagged and the binary release artifacts live at try-foundry/tryfoundry-releases.
How do I report a bug or send feedback?
Inside the app: a feedback button is mounted on every page. Submissions go to the Foundry team and never include your code, file paths, or any project content. Outside the app: email [email protected].
What data does Foundry collect about me?
Anonymous usage telemetry by default (event names + version + OS, no content). Bug reports and feedback when you submit them. Nothing else. See the privacy policy for the full breakdown and how to opt out.