TextGen Goes Portable: Run AI Locally With No Install, No Telemetry

By vramkickedin | | 2 min read
a smooth matte black USB-C flash drive behind it a faint unzipped folder icon made of thin glowing lines.

TextGen is a desktop application that runs large language models locally on your own computer. The latest update transforms the project from a web interface into a no-install portable app for Windows, macOS, and Linux. It handles text, vision, tool-calling, and web search while keeping all data completely offline, and it can act as a local drop-in for OpenAI and Anthropic APIs.

Developer Oobabooga, who started the project in late 2022, has rebuilt it as a self-contained Electron app. The portable package bundles all necessary files, so you just unzip and double-click to start. The free, open-source tool is released under AGPLv3 and accepts no telemetry.

Privacy-first desktop interface

Key Features
  • Portable builds for Windows, Linux, and macOS.
  • Supports CUDA, Vulkan, ROCm, and CPU-only modes.
  • Tool-calling with easy single-file Python integrations.
  • Built-in web search and accurate PDF extraction.
  • 100% offline with zero telemetry or remote requests.
  • Drop-in replacement for OpenAI and Anthropic APIs.
  • Vision support allows image attachments in chats.
  • Create custom characters for role-playing chats.

This tool suits privacy-conscious professionals who need to keep sensitive data local. Hobbyists with prosumer GPUs will appreciate the quick setup and broad hardware support. Small agencies can replace expensive cloud AI subscriptions with a self-hosted alternative that still offers tool-calling and API compatibility, saving costs and keeping data on-premises.

No installation required, just unzip and run

The developer highlighted that TextGen never sends usage data, unlike tools like LM Studio that report hardware details at every launch. It uses ik_llama.cpp for advanced quantization types, offering better model compression than standard llama.cpp. The project remains a passion-driven effort with no plans for commercial licensing.

"Full privacy. Unlike LM Studio, it doesn't phone home on every launch with your OS, CPU architecture, app version, and inference backend choices." — Source: Reddit

Project Links