hkuds/nanobot

Nanobot is an open-source, lightweight personal AI agent that you run on your own machine or server. It sits in the same category as tools like Claude Code or OpenAI Codex, agents that use a large language model as a brain and can autonomously take actions like reading files, running shell commands, browsing the web, and managing tasks, but with a deliberate focus on staying small, readable, and easy to self-host. The problem it solves is that mainstream AI agent platforms either lock you into a proprietary cloud service or require complex infrastructure to deploy. Nanobot aims to let anyone run a capable personal AI agent with minimal setup: install via pip, configure which AI provider to use (it supports Anthropic, OpenAI, DeepSeek, Google Gemini, GitHub Copilot, local models via LM Studio, and many others), and the agent is ready to work. Here is how it works: Nanobot maintains an agent loop where it receives a message or task, reasons about it using the connected AI model, decides which tools to use (file operations, shell commands, web search, notebook editing, etc.), executes those tools, and repeats until the task is complete. It supports MCP (Model Context Protocol) for plugging in additional tools, has a memory system so it can remember context across sessions, and can be reached through multiple channels, Discord, Slack, Telegram, WeChat, Microsoft Teams, email, or a local web UI. Scheduled tasks (cron-style reminders and routines) are also supported. You would use Nanobot if you want a personal AI assistant that can autonomously perform tasks on your computer, respond to you through your preferred messaging app, and be customized or extended without depending on any cloud vendor. The stack is Python (3.11 or later), with direct integrations to multiple AI provider SDKs and chat platform APIs.