Nanobot

Nanobot is an open-source, self-hosted personal AI agent runtime from HKUDS. Instead of a chat interface tied to a specific model or vendor, it gives you a long-running agent process that you own — reachable from a browser WebUI or a terminal, and also wired directly into Telegram, Discord, Slack, WeChat, Email, and Mattermost. The same agent identity shows up wherever you talk to it.

The runtime ships a built-in tool belt (files, shell, web search, web fetch, MCP client, cron scheduling, image generation, and sub-agents) so an agent can do real work rather than just answer questions. Long-term memory is handled by 'Dream' — Nanobot's memory subsystem that turns raw session transcripts into structured recall so agents remember you across restarts. For coordination, the runtime also supports long-horizon goals and scheduled automations that persist across sessions, plus a Python SDK and OpenAI-compatible HTTP API so you can drop it into existing tooling with no code changes.

Nanobot targets people who want a durable, private, always-on AI assistant that runs on their own hardware — a Mac mini, a VPS, or a home server — rather than depending on a hosted chatbot. Because the runtime is model-agnostic (any OpenAI-compatible endpoint, so Claude, GPT, Gemini, DeepSeek, Groq, or local Ollama all work) and MCP-native, it's also a useful base for hobbyists and researchers building custom agent stacks without stitching together LangGraph, CrewAI, and a homegrown chat bridge.