Open-source Character.AI alternatives

SillyTavern is an LLM frontend aimed at power users. It is a web interface for conversing with large language models, and it is built to give people far more control over how conversations are set up than a typical chatbot window provides. The project describes itself with the short tagline of an LLM frontend for power users.

It is widely used for character-driven chat and roleplay, and it connects to a range of text-generation backends, both hosted APIs and locally run models, with extensions that add features. The repository's topics identify it simply as AI, chat and LLM software. Documentation is maintained on a separate site, and there are active Discord and Reddit communities.

SillyTavern is written in JavaScript and licensed under AGPL-3.0, so the source is available and modifications to hosted versions must be shared. You run it yourself on your own computer or server rather than using a hosted service, which keeps your chats and API keys under your control.

MaiBot is an interactive chat agent powered by large language models. Its persona, MaiSaka, is described by the authors as a digital life form rather than a task-oriented assistant, aiming for warmth and authenticity instead of perfection or efficiency, and trying to understand the user and respond in a human way.

A central design point is avoiding the typical assistant voice. The README says replies should read like casual chat of varying length rather than long, markdown-heavy answers. Repository topics list QQ bot support, chatbot and LLM agent keywords, so it is built for messaging platforms. The project is written in Python, has bilingual Chinese and English documentation, and a docs site.

MaiBot is licensed under GPL-3.0 and runs on your own machine or server with your chosen language model. It suits hobbyists and developers interested in persona-driven chatbots, especially for Chinese-language chat communities.

Risuai, often called Risu, is a cross-platform AI chat application and web app aimed at roleplaying and story creation with large language models. It gives users a friendly interface to talk with custom characters, and it is built with TypeScript and Tauri so it can run as a desktop, mobile or browser application.

A large part of its appeal is configurability. It connects to many model providers, including OpenAI, Claude, Gemini, OpenRouter and local backends, and supports group chats with multiple characters, emotion images that change with expressions, lorebooks that give characters longer memory, regex scripts to reshape output, and automatic translation of input and output. Plugins add new features or providers, and prompts can be reordered with conditions and variables.

The project is released under the GPL-3.0 licence. It also includes theme choices, text-to-speech, embedded assets and an MCP client according to its topics. Because it relies on your own provider accounts or local models, it suits hobbyists and writers who want a flexible, free front end for interactive fiction rather than a general-purpose assistant.

TavernAI is a browser-based front end for chatting with, and writing stories alongside, AI language models. Rather than running a model itself, it connects to other backends, and the README lists the Kobold series, Oobabooga's text generation web UI, OpenAI, NovelAI, Claude and Ollama. This repository is the legacy first version; the project points to TavernAI 2 as its successor.

Its interface centers on characters. You can create your own, browse an online character database, and run group chats with several characters at once. Other features include a story mode, world info, swiping between alternative replies, editing, deleting or moving any message, adjustable generation settings, themes (one resembles CharacterAI), custom backgrounds and image recognition with GPT and Claude.

It runs on Node.js with simple start scripts for Windows and Linux, a Windows executable, and a Google Colab option for phones and tablets. The code is JavaScript and licensed under MIT.