Thumper-Run is a desktop application that lets users discover, install, and run open-source AI apps locally on their own hardware, keeping data on-device by default. Offering a catalog of 47 AI tools including ComfyUI, Stable Diffusion, and Open WebUI, the platform provides one-click installation with no terminals, API keys, or usage fees. It supports NVIDIA, AMD, Intel, Apple Silicon, and Qualcomm NPU hardware with auto-detected configuration.
Funding
Funding not disclosed
Founders
Product
Problem
Cloud-based AI services routinely collect user prompts, images, voice data, and other sensitive inputs, which providers may use for training, share with partners, or expose through data breaches. This centralized model forces users to surrender control over their creative process, personal thoughts, business logic, and biometric data to third parties whose terms can change at any time.
Solution
Thumper-Run provides a desktop platform for installing and running open-source AI applications directly on users' own hardware, keeping local work on-device by default. The software auto-detects available compute resources and handles configuration automatically, allowing users to browse a catalog of 47 AI tools and install them with a single click. Local-only workflows operate without an account, network access, or usage fees, while encrypted transport and optional relay-assisted sync provide controlled pathways for workflows that require connectivity. The platform uses open-source model weights that run with no network access unless an app explicitly requires it, with network permissions declared in a manifest visible before installation.
Target Audience
Primary users are privacy-conscious individuals and small teams—including developers, designers, and researchers—who want to run AI tools locally without relying on cloud services or sharing proprietary data with third-party providers.
Features
- One-click installation and launch of 47 AI apps including ComfyUI, Stable Diffusion, and Open WebUI with no terminal commands, API keys, or setup required
- Automatic hardware detection for NVIDIA, AMD, Intel, Apple Silicon, and Qualcomm NPU platforms with zero configuration
- Local-only storage at rest using AES-256-GCM encryption and TLS 1.3 for network transport
- Optional relay-assisted sync with encrypted transport; endpoint-held-key E2EE using BeeKEM/CGKA primitives is a gated design
- App manifests declare network access permissions transparently before installation, preventing unauthorized data transmission
- Data stored in standard formats in user-controlled directories on the local filesystem, ensuring no lock-in