nanobot
nanobot, from the HKU Data Intelligence Lab (HKUDS), is an ultra-lightweight, self-hosted personal AI agent framework written in Python. It runs in a browser WebUI, in the terminal or inside chat apps, and combines tools, long-term memory, MCP integrations, model routing, subagents and scheduled automations around a deliberately small, readable core.
It suits individuals and small teams who want their own always-on assistant without adopting a large platform. Typical uses include chatting with an assistant from Telegram, Discord, Slack, WeChat, Feishu, email or Linear; running long-horizon goals and recurring tasks; and giving the agent tools such as file access, shell, web search, image generation and MCP servers. It works with OpenAI-compatible APIs and local models, offers fallbacks between providers, and exposes a Python SDK and an OpenAI-compatible API for integrations.
Getting started: requires Python 3.11+. Run the one-command installer (or `pip install nanobot-ai`), then `nanobot webui` and choose a provider and model under Settings → Models. `nanobot gateway --background` keeps channels and automations running after you close the terminal.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-11-10
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 3
- Latest release
- v0.4.1 (2026-11-06)
- Open issues
- 874
- Recent issues with at least one comment
- 4/12
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #5266 Logs about token consumption (too many tokens are burned)15 comments · 2026-08-06
- #2463 Architectural issue: nanobot does not preserve the exact prompt prefix it previously sent15 comments · 2026-03-25
- #1955 是否能够或者应该支持在子 agent 执行的过程中,有方式可以看到子 agent 的执行过程?10 comments · 2026-03-13
- #222 Multi agents setup - Supported ?10 comments · 2026-02-06
- #5924 Agent gets stuck in sudo loop - becomes unusuable9 comments · 2026-09-26
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You want a lightweight, hackable Python personal AI agent framework with clean code, a WebUI, and built-in MCP client support.
- You prefer an easy-to-read, self-hosted agent core without the enterprise boilerplate of heavy agent frameworks.
- You require production-grade distributed state checkpointing, strict Pydantic type validation, or multi-role team SOPs; Pydantic AI or CrewAI offer more mature infrastructure.
- Your production stack is .NET or Go; Microsoft Agent Framework covers those runtimes.
How it differs
nanobot provides a full-stack personal agent app with WebUI, memory, and chat tools out of the box. Pydantic AI is a headless, type-safe Python library for embedding model loops into existing services.
nanobot is built as a lightweight personal assistant with a readable core. CrewAI is specialized in role-based multi-agent teams with deterministic Flows and structured task delegation.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!