Swap between GPT-4 and your basement server without leaving the chat

What it does

TavernAI is a browser-based frontend for AI-driven chat and storywriting. It speaks to pretty much every major text generation backend—local engines like KoboldCpp and Ollama, cloud APIs like OpenAI, Claude, and NovelAI, and even the Kobold Horde—so you can run roleplay sessions without pledging allegiance to one model provider. The interface wraps this plumbing in an “atmospheric” skin with configurable backgrounds, themes, and character cards.

The interesting bit

Most frontends in this space pick a backend and build a walled garden; TavernAI treats models as hot-swappable engines behind a single UI. That means you can jump from a basement KoboldCpp instance to GPT-4 mid-conversation, or corral multiple AI characters into a single group chat. It is fundamentally glue code, but it is glue code that remembers your lorebook and lets you retroactively edit any message to keep the narrative on track.

Key highlights

• Backend agnostic: KoboldAI/Horde, Oobabooga, Ollama, OpenAI, Claude, NovelAI
• Group chat with multiple AI characters talking at once
• World info and story mode for managing long-form narrative context
• Message swiping and full edit/delete/move history to iterate on replies
• Visual theming including a CharacterAI-style skin, plus GPT/Claude picture recognition

Verdict

Worth a look if you want a single UI for roleplaying across local and cloud language models. Skip it if you need a headless API or a backend that actually runs the models itself.