// the find
langbot-app/LangBot
Production-grade platform for building agentic IM bots - 生产级多平台智能机器人开发平台/ Agent、知识库编排、插件系统 / Bots for Discord / Slack / LINE / Telegram / WeChat(企业微信, 企微智能机器人, 公众号) / 飞书 / 钉钉 / QQ / Matrix e.g. Integrated with ChatGPT(GPT), DeepSeek, Dify, n8n, Langflow, Coze, Claude, Gemini, MiniMax, Ollama, SiliconFlow, Moonshot, openclaw / hermes agent, deerflow
LangBot is a Python platform for deploying LLM-powered bots across IM platforms — Discord, Telegram, Slack, WeChat, QQ, DingTalk, and about a dozen others. It handles the platform adapter layer, conversation pipelines, RAG knowledge bases, and plugin execution so you don't have to write the glue code for each platform separately. The target user is a developer or small team that wants a self-hosted bot infrastructure without building the multi-platform abstraction from scratch.
Platform breadth is real and maintained — 10+ adapters with their own API client libs bundled under src/langbot/libs/, not just stubs. The migration system is structured (43+ config migrations, separate DB migrations via Alembic), which means upgrades between versions are tracked rather than 'just update the config manually'. The pipeline architecture lets you run different bot configurations per channel or use case without spinning up separate processes. MCP protocol support means tools built for the Model Context Protocol ecosystem plug in without custom code.
The config migration chain (43 files and counting) is a red flag — it means the config schema has been redesigned repeatedly and adopters on older versions face a long upgrade path of incremental patches. WeChat personal account support relies on wechatpad_api, a third-party protocol reverse-engineering library, which is technically a ToS violation and can break without warning when WeChat pushes client updates. The plugin ecosystem lives on an external marketplace with no pinned versions, so a plugin update can silently break your bot. Test coverage appears thin relative to the scope — the test scripts and CI suggest unit tests exist but integration test coverage of the platform adapters is unclear, meaning regressions in adapters probably surface in production first.