nanobot
A self-hosted AI agent for macOS, Windows and Linux, with Ollama and vLLM support, persistent memory, chat app connections and an MIT license.
nanobot is a self-hosted personal AI agent for people who want an assistant that can work with files, carry tasks across conversations and handle recurring jobs. It runs on macOS, Windows or Linux, on your own server, or in Docker. Its Python core is open source under the MIT license and small enough for developers to inspect and extend.
You can use local LLM backends such as Ollama and vLLM, or connect to cloud providers through OpenAI-compatible APIs. The agent runs wherever you host it; choosing a cloud model sends model requests to that provider. Render deployment is also available for hosting the agent and browser interface away from your machine.
The browser and terminal interfaces share saved conversations. The browser lets you keep project topics side by side, switch models and workspaces, and inspect tool calls, file diffs, command output and generated artifacts. Temporary chats stay out of saved history and long-term memory.
For work beyond chat, nanobot can read and edit files, execute shell commands, search the web and generate images. MCP connections extend its tools, while reusable Skills supply task instructions. It can delegate work to subagents, retain long-term memory through Dream, and pursue goals across extended tasks.
Scheduled automations and local triggers support recurring or on-demand jobs. Connections to Telegram, Discord, Slack, WeChat, email and Linear let you interact through existing communication tools. A Python SDK and an OpenAI-compatible API let other software call the agent.
