Everything people actually ask us.

A few cents per resume with most providers, paid directly to them at their rates. Or skip keys entirely and run a local Ollama model for free.

Yes. Point it at a local Ollama model and every AI call happens on your machine, with no network request and no account.

Only the text you send to whichever AI provider you chose, the same as pasting into ChatGPT. Your profile, jobs, and resumes live in a local SQLite database that Udaan never uploads.

Yes. The PDF export uses standard text layout instead of images or exotic fonts, which is what most ATS parsers choke on.

It rewrites your base profile rather than inventing one. Every AI edit is a suggested change you review before it's applied, so nothing exports without you seeing it first.

Udaan is self-signed rather than paid-publisher signed, so macOS and Windows show a one-time warning for any app outside their stores. See the note below the download button for the two-click fix.

It ranges from 62MB on Windows to 179MB for the Linux AppImage, with the macOS .dmg around 113MB. Tauri keeps the app binary itself small, but Udaan bundles a full Next.js server to run its local API and database layer, and that server is most of the weight. It's still a native window rather than a wrapped browser tab, so startup and memory use stay closer to Tauri than Electron.

macOS, Windows, and Linux, as of v1.13.1. That release was the first to build all three: separate Apple Silicon and Intel .dmg files, an .exe installer and .msi for Windows, and .AppImage, .deb, and .rpm for Linux. Anything older than v1.13.1 is macOS only. There's no universal macOS binary, so pick the one matching your chip.

On desktop, in a file encrypted with AES-256-GCM, where the encryption key comes from a per-install master key held in your OS keychain. On the web build, they sit in localStorage. Either way, they never touch Udaan's SQLite database or leave your machine.

Settings has a Backup & Restore page that exports your full database as JSON. It deliberately excludes your API keys, so restoring on a new machine still asks you to re-enter those.

An update replaces the app bundle but never touches your existing database. On launch, a migration step alters your database in place to match the new schema, so it's a no-op once you're already current.

German and EU conventions are the default, not an add-on. Your profile carries optional fields for a photo, nationality, date of birth, work authorization, and hobbies, and there's a Euro Sidebar template plus an Anschreiben cover-letter style. When the job is EU-based, the tailoring follows German rules: two pages rather than a crammed one, no first-person pronouns, CEFR language levels, no skill bars. Paste a US posting and you get a US résumé instead.

You can write and edit a resume by hand start to finish, with PDF/TXT export, templates, and the deterministic proofread checks all working with no model configured. AI-specific features like tailoring and the chat assistant need a provider, but they're optional.

It exposes Udaan's own flows as tools Claude can call directly: parsing a job description, listing your jobs and profiles, generating and editing a tailored resume, and exporting files. Claude runs the conversation on the subscription you already pay for; Udaan only reads and writes your local database.

The app itself has no telemetry. This marketing site runs a privacy-focused analytics beacon (page views and Core Web Vitals, no cookies, no personal data), but the desktop app you actually use does not.