Fully production-grade implementation of AI-Agent-Nanobot, refactored by the 69-Agent Opencode Collective.
- CI/CD: Automated GitHub Actions with Gitleaks security scans.
- Security: Standardized SECURITY.md protocol.
- Design: Opencode Premium Design Tokens integrated.
- Initialize infrastructure via
.github/workflows. - Set your secrets in GitHub Environment settings.
- Deploy to production via Vercel/Docker.
~4,000 lines. Infinite possibilities. The ultra-light personal AI lab.
🌐 Live Website · 📖 Docs · 🚀 Quick Start
Nanobot is a minimalist but powerful AI agent built in Python. It delivers 99% of the functionality of larger frameworks like OpenClaw with a fraction of the code complexity. Designed to run on everything from a Raspberry Pi Zero to a high-end cloud VM.
- Ultra-Lightweight: Core agent logic is under 4,000 lines of readable Python.
- Multi-Channel: Native support for Telegram, Discord, Slack, WhatsApp, and Email.
- MCP Integration: Full support for Model Context Protocol (MCP) toolkits.
graph LR
subgraph Input["9 Channels"]
TG[Telegram]
DC[Discord]
SL[Slack]
WA[WhatsApp]
FE[Feishu]
WC[WeChat]
EM[Email]
WB[WebChat]
SS[Signal]
end
subgraph Core["nanobot Core (~4,000 lines)"]
RT[Message Router]
AG[Agent Loop]
MM[Memory Manager]
HB[Heartbeat]
PC[Prompt Cache]
end
subgraph Providers["22+ LLM Providers"]
CL[Claude / Anthropic]
OR[OpenRouter]
DS[DeepSeek Coder]
VL[Local vLLM]
OA[OpenAI]
end
subgraph Tools["MCP Tools"]
GH[GitHub API]
SB[Supabase]
BS[Brave Search]
FS[Filesystem]
end
Input --> RT
RT --> AG
AG --> MM
AG --> HB
AG --> Providers
AG --> Tools
Nanobot's power comes from its modular skills. Each skill is designed for high-fidelity task execution.
| Skill | Description | Status |
|---|---|---|
web-searcher |
Brave Search + AI synthesis. Performs live research with citations. | 🛡️ New |
notion-sync |
Two-way sync for tasks, notes, and research into Notion. | 🛡️ New |
code-reviewer |
Reads git diff, analyzes quality, checks security, auto-comments PRs. Uses DeepSeek Coder via OpenRouter | ✅ Active |
git-automator |
Automated branching: feature branches, PR management, auto-merge on CI pass, changelog generation | ✅ Active |
data-pipeline |
CSV/JSON ingestion, clean/transform, analysis reports, visualizations on scheduled cron | ✅ Active |
api-tester |
Reads OpenAPI spec → generates test cases → runs automated tests → reports with latency metrics | ✅ Active |
doc-generator |
Scans Python/TypeScript codebases, extracts docstrings + types, generates Markdown API docs | ✅ Active |
Automate complex pipelines with simple Markdown definitions.
| Workflow | Trigger | Output |
|---|---|---|
hr-campaign |
Autonomous HR outreach pipeline. Scrapes leads, personalize drafts, and rotates sending. | ✅ Active |
mcp-tool-chain |
GitHub API + Supabase + Brave Search orchestrated in one session | ✅ Active |
multi-channel-bot |
Single nanobot instance across 9 channels with scoped permissions | ✅ Active |
clawwork-coworker |
Professional task completion, token cost tracking, income reports | ✅ Active |
memory-optimization |
Tune memory for low-RAM devices (RPi Zero, SBC) | ✅ Active |
How I use Nanobot daily:
Autonomous expense tracking via Telegram/WhatsApp. Categories spending from receipt photos and transaction logs.
- Docs: Finance Monitor Use Case
A morning briefing (news + calendar) and real-time task manager running on 24/7 edge hardware.
A specialized system using REPO_UPGRADE_TRACKER.md to manage the evolution of 80+ repositories.
Coding help bot deployed in my Discord server:
- Answer questions about my tech stack (Next.js, FastAPI, Supabase)
- Review code snippets in threads
- Generate boilerplate from descriptions
- Run quick API tests on demand
AI-Agent-Nanobot/
├── config/ # nanobot config templates
│ └── config.example.json
├── docs/ # Deployment & Provider guides
│ ├── deployment.md # SBC & Cloud setup
│ ├── providers.md # LLM config (Claude, OpenRouter)
│ └── REPO_UPGRADE_TRACKER.md # Portfolio management system
├── providers/ # Provider setup guides
│ ├── openrouter-setup.md
│ ├── local-vllm.md
│ └── deepseek-coder.md
├── skills/ # High-fidelity skill definitions
│ ├── web-searcher/ # Live search engine
│ ├── notion-sync/ # Notion workspace integration
│ ├── code-reviewer/ # Logic analysis
│ ├── git-automator/
│ ├── data-pipeline/
│ ├── api-tester/
│ └── doc-generator/
├── use-cases/ # Real-world deployment examples
│ ├── personal-finance-monitor/
│ ├── telegram-productivity-bot/
│ ├── discord-code-assistant/
│ └── clawwork-coworker/
├── workflows/ # Automation pipelines
│ ├── hr-campaign.md # HR outreach system
│ ├── mcp-tool-chain.md
│ ├── multi-channel-bot.md
│ ├── clawwork-coworker.md
│ └── memory-optimization.md
├── website/ # Next.js → Vercel
└── assets/ # Premium branding & assets
# Install from PyPI
pip install nanobot-ai
# Or from source
git clone https://github.com/HKUDS/nanobot.git
cd nanobot
pip install -e .
# Configure
cp config.example.json config.json
# Edit: model, channels, API keys
# Run
python -m nanobot
# Verify line count
bash core_agent_lines.sh- Multi-channel deployment (9 channels)
- OpenRouter provider routing
- MCP tool chain integration
- Heartbeat redesign (v0.1.4)
- Local vLLM guide for RPi
- Cost tracking dashboard
- Nanobot → ZeroClaw migration guide
- Community skill registry
| Repo | Focus |
|---|---|
| AI-Agent-OpenClaw | 🦞 Full-stack Hub |
| AI-Agent-Nanobot | 🐈 Lightweight Lab · ← You are here |
| AI-Agent-ZeroClaw | 🦀 Rust Runtime |
| AI-Agent-PicoClaw | 🦐 Edge/IoT |
| AI-Agent-NanoClaw | 🐚 Swarm Agent |
Part of the Claw Ecosystem by mk-knight23
MIT © mk-knight23
This project follows security best practices:
- No hardcoded credentials
- Dependency scanning enabled
- Security headers configured
- Regular security audits performed