Nanobot

⚠️ This template requires a Dedicated Server on Zeabur and cannot run on shared clusters.

Recommended resources: 1 vCPU / 2 GB RAM (minimum: 0.5 vCPU / 256 MB RAM)

Nanobot is an ultra-lightweight personal AI assistant framework (~4K lines of core agent code) from HKUDS. It supports multiple LLM providers and connects to various messaging platforms via a unified Gateway architecture.

Official documentation: github.com/HKUDS/nanobot

Supported Chat Platforms

Telegram, Discord, WhatsApp, Slack, DingTalk, Feishu, QQ, and Email.

Quick Start

1. After deployment, the Nanobot gateway starts automatically
2. If you entered an API key during deployment (Zeabur AI Hub or OpenRouter), a chat model is already configured
3. If you entered a Telegram bot token, the bot is ready — send it a message to start chatting
4. To add more platforms or change settings, edit /root/.nanobot/config.json via the Files tab

Skipped a field during deployment? You can add it later by editing /root/.nanobot/config.json in the Files tab, or by adding the corresponding environment variable in the Zeabur dashboard and restarting the service.

Default model: When using Zeabur AI Hub, the default model is openai/claude-sonnet-4-5. When using OpenRouter, the default model is openrouter/anthropic/claude-sonnet-4. You can change this in the config file at any time.

LLM Providers

Nanobot supports many providers. You can switch at any time by editing /root/.nanobot/config.json via the Files tab and restarting the service.

Provider	How to get API key	providers config	Model ID example
Zeabur AI Hub	zeabur.com/ai-hub	"openai": {"api_key":"...","api_base":"https://hnd1.aihub.zeabur.ai/v1"}	openai/claude-sonnet-4-5
OpenRouter	openrouter.ai/keys	"openrouter": {"api_key":"sk-or-..."}	openrouter/anthropic/claude-sonnet-4
Anthropic	console.anthropic.com	"anthropic": {"api_key":"sk-ant-..."}	anthropic/claude-sonnet-4-5
OpenAI	platform.openai.com	"openai": {"api_key":"sk-..."}	openai/gpt-4o
DeepSeek	platform.deepseek.com	"deepseek": {"api_key":"..."}	deepseek/deepseek-chat

To switch provider:

1. Open /root/.nanobot/config.json in the Files tab
2. Add or update the provider entry under "providers"
3. Change "agents" > "defaults" > "model" to the new model ID
4. Restart the service for changes to take effect

Example — switch from Zeabur AI Hub to OpenRouter:

{
  "providers": {
    "openrouter": { "api_key": "sk-or-your-key" }
  },
  "agents": {
    "defaults": { "model": "openrouter/anthropic/claude-sonnet-4" }
  }
}

Chat Platforms

Supported: Telegram, Discord, WhatsApp, Slack, DingTalk, Feishu, QQ, and Email. See the official documentation for the full list and setup guides.

Telegram setup:

1. Search for @BotFather on Telegram and send /newbot
2. Follow the prompts to set the bot name and username
3. Copy the bot token you receive
4. Enter it during deployment, or add it to config later

If you entered the bot token during deployment, it's already active. Otherwise, add it to config:

"channels": {
  "telegram": { "enabled": true, "token": "BOT_TOKEN", "allow_from": [] }
}

Other platforms — add the corresponding channel in config and restart:

"channels": {
  "discord": { "enabled": true, "token": "BOT_TOKEN", "allow_from": [] },
  "slack": { "enabled": true, "bot_token": "xoxb-...", "app_token": "xapp-...", "allow_from": [] }
}

allow_from — controls who can talk to the bot. Empty list [] = anyone can use it. To restrict access, add user IDs (for Telegram, send a message to @userinfobot to find your user ID):

"allow_from": ["123456789", "987654321"]

Restart the service after any config changes.

CLI Commands

Access the terminal via Command in Zeabur dashboard:

• nanobot agent -m "query" - Send a single message
• nanobot agent - Interactive chat
• nanobot status - View provider configuration

Data Persistence

All data is stored under /root/.nanobot (mounted as a persistent volume):

• config.json - Configuration
• workspace/ - Agent workspace and memory

It is recommended to use Zeabur's backup feature to regularly back up your data.

Troubleshooting

Service won't start?

1. Check Logs for error messages
2. Edit config via Files tab to fix issues
3. Restart the service

Update version: Modify the startup script to install a specific version: uv pip install --system nanobot-ai==0.1.3.post6

More help: See the official documentation for detailed configuration and usage guides.