HKUDS/nanobot

Nanobot is a personal AI agent framework that runs as a self-hosted Python application with a WebUI, terminal interface, and chat app integrations.

The framework addresses the need for a lightweight, self-contained AI agent system by combining several capabilities in a single core: tool execution, long-term memory persistence, Model Context Protocol integrations, model routing across different LLMs, multi-agent workflows with delegation, and scheduled automation. It exposes an OpenAI-compatible API, allowing it to work with existing tooling and clients designed for that interface. The codebase is designed to be small and readable, making it accessible for developers who want to understand or modify their AI agent infrastructure.

Nanobot suits developers and individuals who want to self-host their AI agent infrastructure without external dependencies, prefer Python-based systems they can inspect and extend, or need to run agents locally with privacy guarantees. It works well for personal automation, multi-agent workflows, and scenarios where chat app integration (Discord, Telegram) or scheduled task execution is required. The project positions itself as an alternative to heavier frameworks by emphasizing its lightweight footprint and readable implementation.

The project maintains an active test suite with continuous integration. Documentation is comprehensive and available in multiple languages. The codebase demonstrates consistent maintenance with a focus on stability and accessibility for contributors seeking to understand or extend the framework.