One Go binary to bot them all, from Telegram to DingTalk

What it does MuseBot is a Go-based relay that sits between popular LLM providers and a sprawling list of chat apps—Telegram, Discord, Slack, Lark, DingTalk, WeChat Work, QQ, WeChat, and a generic Web API. It forwards user messages to backends like OpenAI, Gemini, DeepSeek, or Doubao, then streams the AI’s text, image, or video responses back into the conversation thread. An admin dashboard and service-registration module are bundled in, so you can manage multiple instances without editing scattered config files by hand.

The interesting bit Instead of treating each chat platform as a custom integration, MuseBot normalizes them into a single gateway, then layers on RAG, cron-triggered LLM jobs, and MCP-to-function-call translation. The README’s provider matrix is refreshingly blunt: Gemini, Doubao, and Qwen handle the full multimodal stack including video and TTS, while DeepSeek remains text-only—so your bot’s superpowers depend entirely on whose API key is plugged in.

Key highlights

• Normalizes nine chat platforms (Western and Chinese consumer/enterprise) through one Go binary
• Multimodal feature set varies by LLM: text, image generation, video creation, photo recognition, and TTS where the provider exposes it
• Built-in admin platform, metrics endpoint, and service registration for multi-instance ops
• Cron scheduler can fire LLM workflows automatically, no human prompt required
• MCP protocol translated into function calls for tool-using conversations

Caveats

• The capability matrix is honest but uneven: DeepSeek offers only text generation, and several providers lack video or TTS, so feature availability is strictly pay-for-what-your-vendor-allows.
• Documentation is thorough yet fragmented across many small files in static/doc, meaning platform-specific quirks may require some digging.

Verdict A solid fit if you run a mixed fleet of Chinese and Western chat apps and want one self-hosted AI backend behind all of them. If you only need a single-platform bot, lighter purpose-built alternatives will mean less operational surface area.