CaseDeskSupport & download
CaseDesk is in free beta on Windows, Linux and macOS (Apple silicon).
Download
v1.0.49 · Windows 10/11 · ~1.15 GB · digitally signed by Fulks, Inc. No account, no credit card — the app runs entirely on your machine. Questions or want us to know who's testing? Email admin@casedesk.law.
Verify your download — SHA-256:dbd97ae142f057340538d3294fdb1ddb600a838a6e0607ca810c69ff3d86604c
First run: Windows SmartScreen may show “Windows protected your PC” while a new build earns reputation. Click More info, then Run anyway, and confirm the publisher reads Fulks, Inc.
macOS: Apple silicon .dmg (~317 MB, M1 and later — there is no Intel build). v1.0.49 · signed and notarized by Apple, so it opens without a Gatekeeper warning. Local AI runs on Metal. On first launch macOS may ask for permission to access folders you point CaseDesk at; the app stores everything on your machine.
SHA-256 (.dmg): 684b0c83e9f46fb97503124c1843513d82e930fd4ae9ac98872e8c75d1f074f9
Phone: app.casedesk.law — the phone companion, opened in your phone’s browser and added to the home screen. It reads your end-to-end-encrypted Cloud Sync, so it needs CaseDesk Pro with Cloud Sync on in the desktop app; join with your Cloud Sync recovery code. The iPhone and Android store apps are the same companion and are in App Store / Google Play review.
Linux: .deb for Ubuntu/Debian (~808 MB, recommended — it installs the AppArmor profile the Chromium sandbox needs on Ubuntu 24.04+) or AppImage for other distros (~945 MB, needs libfuse2; ⚠ on Ubuntu 24.04 and newer, and any distro that sets kernel.apparmor_restrict_unprivileged_userns=1, the AppImage will not start on a plain double-click — those kernels block the user-namespace sandbox and an AppImage cannot carry the setuid helper Chromium falls back to. Run it as ./CaseDesk-1.0.49.AppImage --no-sandbox, or use the .deb, which ships a root-owned sandbox helper and needs no workaround). v1.0.49 · 64-bit · the same free beta as Windows, and the newest platform — released this month, so it has less field mileage. Local AI included, and it runs on your graphics card: NVIDIA goes through CUDA, AMD and Intel through Vulkan, with an optional extra download for Intel Arc in Settings. If you use automatic login, sign in with your password once — Linux leaves the login keyring (which guards the database encryption key) locked until a password login.
SHA-256 (.deb): cf613d951435c193528d299842b43a78b70469d7efa25e8fb4b0350e502d402e
SHA-256 (AppImage): a62021ef24b33693cdc38f0aac53d8ee56301590983801ba1419c2e0d21296df
System requirements
The app itself is modest. Local AI is the only part with real hardware appetite, and it is entirely optional.
| Operating system | Windows 10 or Windows 11, 64-bit, or Linux (Ubuntu/Debian via .deb, other 64-bit distros via AppImage) — both platforms are in the same free beta, Linux being the newer of the two. macOS is not available during the beta. |
|---|---|
| Memory | 8 GB RAM for everything except local AI. |
| Disk | About 3 GB for the application. Your case database and documents grow separately, wherever you put them. |
| Internet | Not required. The whole app works offline. Apart from checking our download host for a new version, connections happen only for integrations you turn on. |
| AI — cloud key | No special hardware. Bring your own Anthropic (Claude) or OpenAI key — or point CaseDesk at any OpenAI-compatible endpoint you run or subscribe to — and it talks to that provider directly from your machine. |
| AI — the models | The local catalog: DeepSeek V4 Flash (multi-GPU workstation flagship, up to a 1M-token context), Qwen 3.8 27B (three sizes, 262K context, plus a 512K extended-context option), Qwen 3.6 27B (with multi-token-prediction speed-up where video memory allows), Qwen 3.6 35B-A3B mixture-of-experts, Gemma 4 31B, Gemma 4 26B MoE, and Gemma 4 E4B/E2B for small GPUs and CPU-only machines — most in two sizes. CaseDesk recommends the right one for your hardware, and you can also point it at any GGUF model file of your own. |
| AI — local, no GPU | Runs a small model on the CPU. Usable for structured tasks; slower for long drafting — expect the first answer in a conversation to take a few minutes while the model reads your context. |
| AI — local, 6–8 GB GPU | A graphics card in this range runs a solid mid-size model at comfortable speed. NVIDIA, AMD and Intel cards are all accelerated, on Windows and on Linux. |
| AI — local, 20–24 GB GPU | Runs the strongest local model CaseDesk ships a profile for — the one worth using for legal reasoning and long drafting. |
| AI model download | Separate and one-time, from about 3.5 GB up to 145 GB depending on which model you choose; most people run one between 5 and 20 GB. CaseDesk detects your hardware and recommends one; it is not bundled in the installer. |
Why the installer is ~1.15 GB: most of it is the local-AI inference engine, shipped in the box so private AI works without contacting a server. The legal library, court database, deadline rules, and templates are in there too.
Getting help
Same address for questions, bug reports, and feature requests. CaseDesk is built and supported by a practicing attorney, so replies may take a business day; include screenshots and the steps that led to the problem when reporting a bug.
Your data, in plain terms
Support never requires sending us your case data — it stays on your computer. If a bug report needs technical detail, we'll tell you exactly which non-privileged log or setting to look at.
Privacy & terms
See the Privacy Policy and Terms of Use.