TextGen (text-generation-webui)
oobaboogaTextGen is a local application for running large language models, usable as a desktop window or a browser interface. It loads models through llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3 and TensorRT-LLM, serves OpenAI- and Anthropic-compatible endpoints with tool calling and MCP servers, and adds LoRA training and a diffusers image-generation tab. The project, by the developer oobabooga, was long known as text-generation-webui.
The repository was renamed from oobabooga/text-generation-webui to oobabooga/textgen; the slug keeps the older name.
Openness
5 high confidence- license
- AGPL-3.0 (OSI. the LICENSE body is the GNU Affero General Public License version 3)
- source
- public (oobabooga/textgen)
- core features withheld
- no — no ee, enterprise, commercial or proprietary path. a project by an individual developer, with no company, paid edition or hosted service named
TextGen is released under the GNU AGPL version 3 with its full source public, and its desktop builds are made from the same repository. It is one developer's project, with no company, paid edition or hosted service beside it.
- https://raw.githubusercontent.com/oobabooga/textgen/main/LICENSE recorded 2026-11-12
LICENSE body: "GNU AFFERO GENERAL PUBLIC LICENSE / Version 3, 19 November 2007", the stock text with no added terms.
- https://raw.githubusercontent.com/oobabooga/textgen/main/README.md recorded 2026-11-12
README: "A desktop app for local LLMs. Open source, no telemetry."; portable builds come from the repository's releases page, and the manual install clones this repository and runs server.py. No paid tier or hosted service is described.
- https://ungh.cc/repos/oobabooga/textgen/files/main recorded 2026-11-12
Full file list of the main branch (modules/, extensions/, desktop/, docker/, requirements/): no ee/, enterprise/, commercial/ or proprietary/ path.
Adoption
3 low confidenceTextGen ships as portable builds and a cloned repository rather than through a package registry, so GitHub stars are the only signal. A star is not a use, so the reading says little about how many people run it.
- https://ungh.cc/repos/oobabooga/textgen recorded 2026-11-12
Repository record for oobabooga/textgen (name "textgen"): stars 47737, pushedAt 2026-08-17.
Capability
4 medium confidenceTextGen runs local models on five inference backends and serves them through OpenAI- and Anthropic-compatible endpoints with tool calling and MCP, on par with LM Studio. It is a single-user application without the multi-user accounts and agent tooling that put LibreChat higher.
- https://raw.githubusercontent.com/oobabooga/textgen/main/README.md recorded 2026-11-12
README Features: "Multiple backends: llama.cpp, ik_llama.cpp, Transformers, ExLlamaV3, and TensorRT-LLM"; "OpenAI/Anthropic-compatible API: Chat, Completions, and Messages endpoints with tool-calling support"; "MCP servers are also supported"; "Training: Fine-tune LoRAs"; "Image generation: A dedicated tab for diffusers models".
Verified 2026-11-12