Docs · Guide

nanobot on any model

nanobot (HKUDS/nanobot, pip install nanobot-ai) is a personal AI assistant that answers in chat apps, its CLI and its WebUI. Its model comes from ~/.nanobot/config.json, where any key under providers that isn't one of its own is an OpenAI-compatible endpoint. magpie adds a magpie provider there and sets the default model to one of magpie's, so nanobot runs on any model you have in magpie: DeepSeek, GLM, Kimi, Qwen, an OpenAI or Anthropic key, or a ChatGPT or Copilot subscription.

Set up nanobot

1. Run nanobot oncenanobot onboard or its first run makes ~/.nanobot.
2. Add a provider in magpieAdd an API key on the Providers page, or sign in to a subscription.
3. Pick a model on nanobot's rowOn the Agents page, pick a magpie model for nanobot. A nanobot in WSL is set up there too.

pip install nanobot-ai && nanobot onboard   # nanobot makes ~/.nanobot
magpie nanobot deepseek/deepseek-v4-pro

What goes in config.json

magpie adds providers.magpie at the gateway's /v1, with the gateway's key and a User-Agent of nanobot (the OpenAI SDK's otherwise, which tells it from no one), and sets agents.defaults to provider magpie, model magpie/<model id> and the context window magpie knows the model to take. nanobot strips the magpie/ prefix before it sends the model.

// ~/.nanobot/config.json
{
  "providers": {
    …,
    "magpie": {"apiBase": "http://127.0.0.1:3425/v1", "apiKey": "…",
               "extraHeaders": {"User-Agent": "nanobot"}}
  },
  "agents": {"defaults": {
    "provider": "magpie",
    "model": "magpie/deepseek/deepseek-v4-pro",
    "contextWindowTokens": …
  }}
}

nanobot's pydantic models read camelCase and snake_case keys alike. magpie writes nanobot's own camelCase, and takes out a snake_case spelling of the same key, so no two settings disagree.

modelPreset

When agents.defaults.modelPreset names one of modelPresets, it goes over both provider and model. So magpie stashes it and takes it out while magpie's model is on, and puts it back when it steps out; your modelPresets stay as they are.

When it takes effect

nanobot's gateway (nanobot gateway, the WebUI's) watches the file and takes a change at its next reply. nanobot agent reads it once, so restart an open nanobot agent chat. magpie says so while nanobot is running.

Your own settings

The provider, model, context window and modelPreset you had under agents.defaults are stashed and put back when magpie steps out. Your other providers, presets and keys stay as they are. A config.json magpie made goes again once nothing is left in it. If you delete the magpie provider in nanobot, magpie's check says it no longer reaches magpie.

Switching back

Set the row back to default in magpie: magpie takes out providers.magpie and puts agents.defaults back. Or pick a model of your own, and magpie puts things back and then sets it as the model.

magpie nanobot default

FAQ

Can nanobot use DeepSeek, GLM, Kimi or Qwen?

Yes. magpie adds an OpenAI-compatible magpie provider to ~/.nanobot/config.json and sets the default model to the one you pick in magpie.

Why is nanobot still on the old model?

nanobot's gateway and WebUI take the change at their next reply; nanobot agent reads its config once, so restart the chat.

Do I lose my modelPreset?

No. magpie stashes agents.defaults.modelPreset and takes it out (it would go over magpie's model), and puts it back when it steps out; modelPresets stay.

Does it work for nanobot in WSL?

Yes. magpie sets up a nanobot in a WSL distro too.

More agents on the same models: OpenClaw on any model, Hermes Agent on any model, OpenChamber on any model and AstrBot on any model. Something fails? See Fixing common errors, or ask on Discord.