llmcord
A tool that turns Discord into a collaborative front end for any OpenAI-compatible LLM, using reply chains as conversation history so no database is required.
llmcord turns a Discord server into a shared LLM chat interface: mentioning the bot starts a conversation, and replying to a message continues it, with the reply chain itself serving as conversation history so no database is needed. Conversations can branch endlessly, be continued by other users, or be moved into a thread, and back-to-back messages from the same user are chained together automatically.
It works with remote models from OpenRouter, OpenAI, xAI and Google, or local models through LM Studio, Ollama and vLLM, and switching models is done with a /model command. It supports image attachments for vision models and text file attachments, a customizable system prompt, streamed responses, and hot-reloading configuration, and is implemented in about 300 lines in a single Python file.