The history and evolution of Nanobot, an ultra-lightweight Python AI agent inspired by OpenClaw.
Nanobot is an ultra-lightweight personal AI assistant that delivers core agent functionality in just ~4,000 lines of code — 99% smaller than Clawdbot’s 430k+ lines. The project was created to provide a clean, readable, and portable alternative to complex AI agent frameworks.
Nanobot emerged in early 2026 as a response to the growing complexity of AI agent frameworks. While projects like OpenClaw (Clawdbot) offered features, their 430,000+ lines of code made them difficult to understand, modify, and deploy on resource-constrained hardware.
The Nanobot creators asked: “What if we could deliver 90% of the value with just 1% of the code?”
Nanobot is developed by HKUDS (Hong Kong University Data Science, Data Intelligence Lab) and individual contributors. The project is maintained as an open-source initiative under the MIT License.
- Initial Concept - Design ultra-lightweight agent architecture
- Code Philosophy - Prioritize readability and simplicity
- Platform Support - Target 10+ messaging platforms from day one
- v0.1.0 - Initial public release
- Core Features - Telegram, Discord, WhatsApp support
- PyPI Package - Published as
nanobot-ai
- Docker Support - Dockerfile and docker-compose.yml
- v0.1.3 - Added Slack, Feishu, DingTalk support
- v0.1.3.1 - Discord file support, message splitting
- v0.1.4 - MCP (Model Context Protocol) support
- v0.1.4.post3 - February 28, 2026
- v0.1.4.post4 - March 8, 2026 - Authorization bypass fix (GHSA-g6cf-xjwr-pm77), 58 PRs merged, 29 new contributors
- v0.1.4.post5 - March 16, 2026 - Security hardening, multi-instance improvements, 57 PRs merged, 29 new contributors
- v0.1.4.post6 - March 27, 2026 - Removed litellm dependency, native SDKs, end-to-end streaming, WeChat support, security fix (GHSA-4gmr-2vc8-7qh3 / CVE-2026-33654)
- v0.1.5 - April 6, 2026 - Major release with official website (nanobot.wiki), memory architecture overhaul (“Dream” system), production hardening (bwrap sandbox, non-root containers), GPT-5 support, 66 PRs merged, 27 new contributors
| Date |
Milestone |
| January 2026 |
Project started |
| February 2026 |
v0.1.0 public release |
| February 2026 |
1,000 GitHub stars |
| February 2026 |
v0.1.3 with additional platform support |
| February 2026 |
10,000 GitHub stars |
| February 2026 |
v0.1.3.post7 security fix (CVE-2026-2577) |
| February 2026 |
26.9k GitHub stars |
| March 2026 |
v0.1.4.post4 release (security fix GHSA-g6cf-xjwr-pm77) |
| March 2026 |
v0.1.4.post5 release (security hardening) |
| March 2026 |
v0.1.4.post6 release (native SDKs, WeChat, security fix) |
| March 2026 |
35.7k+ GitHub stars |
| April 2026 |
v0.1.5 release (major update, nanobot.wiki launched) |
| April 2026 |
38.7k+ GitHub stars, 6.8k+ forks |
Nanobot’s defining characteristic is its minimal codebase:
| Project |
Lines of Code |
Relative Size |
| Clawdbot |
430,000+ |
100% |
| Nanobot |
~4,000 |
~1% |
This 99% reduction enables:
- Faster startup and execution
- Easier understanding and modification
- Lower resource requirements
- Raspberry Pi-friendly deployment
Python-First Design:
- 97.0% Python codebase
- Leverages Python’s extensive ecosystem
- Easy to extend and modify
Minimal Dependencies:
- Core agent with minimal external dependencies
- Optional extras for specific platforms (e.g., Matrix support)
- WhatsApp bridge in Node.js (only when needed)
Configuration Simplicity:
- Single JSON configuration file
- No complex environment variable management
~/.nanobot/config.json for all settings
Initial Release (v0.1.0):
- Telegram
- Discord
- WhatsApp (via Node.js bridge)
v0.1.3 Additions:
- Slack (Socket Mode)
- Feishu (飞书)
- DingTalk (钉钉)
Later Additions:
- QQ (via botpy SDK)
- Email (IMAP/SMTP)
- Matrix (with E2EE support)
- Mochat (Claw IM)
| Metric |
Value |
| Stars |
38,700+ |
| Forks |
6,800+ |
| Language |
Python (98.6%) |
| License |
MIT |
| Releases |
v0.1.5 (latest) |
| Status |
Active development |
| Documentation |
nanobot.wiki (7 languages: EN, ZH, JA, KO, ES, FR) |
¶ Community and Ecosystem
Nanobot exists in the same ecosystem as OpenClaw:
- OpenClaw - Full-featured AI agent
- Nanobot - Ultra-lightweight, focused on simplicity
- ZeroClaw - Rust rewrite for performance
All three projects share the goal of accessible, self-hosted AI agents.
| Project |
Description |
| OpenClaw |
Full-featured AI agent with persistent memory |
| ZeroClaw |
Lightweight Rust rewrite of OpenClaw |
| NanoClaw |
Apple Container-based agent swarms |
| Moltis |
Rust single-binary agent runtime |
Nanobot is released under the MIT License:
- ✅ Free to use, modify, and distribute
- ✅ Commercial use allowed
- ✅ No warranty provided
- ✅ Copyright notice required
This permissive license has contributed to rapid adoption and community contributions.
Nanobot has influenced the self-hosted AI space by:
- Demonstrating Simplicity - Proving that less code can deliver core value
- Lowering Barriers - Making AI agents accessible on low-power hardware
- Encouraging Learning - Clean codebase for educational purposes
- Inspiring Alternatives - Sparking discussion about minimalism in AI tools
- Latest Version: v0.1.5 (April 6, 2026)
- Release Cadence: Multiple updates per week
- Community: 38.7k+ stars, 6.8k+ forks
- Package: Available on PyPI as
nanobot-ai and nanobot-ai-tng
- Documentation: nanobot.wiki — multilingual (EN, ZH, JA, KO, ES, FR)
- Code Size: ~4,000 lines of code (dynamically verified via
bash core_agent_lines.sh)
Major release with 66 PRs merged and 27 new contributors:
- 🌐 Official Website — Launched nanobot.wiki with multilingual documentation (EN, ZH, JA, KO, ES, FR)
- 🔒 Long-Running Task Reliability — Ground-up reliability pass for extended sessions; handles
CancelledError without orphaning subprocesses, structured retry metadata, disables SDK-level auto-retries to prevent request amplification, fixes Azure retry duplication
- 🧠 Memory Architecture (“Dream”) — New two-stage system separating live conversation history from consolidated long-term knowledge; background consolidation, git-versioned storage, automatic legacy
HISTORY.md migration, Jinja2 templating
- 🛡️ Production Hardening — Exec sandboxed via
bwrap, containers run as non-root by default, API port binds to localhost, ${VAR} secret interpolation, new nanobot-api Docker service for isolated OpenAI-compatible endpoints, WhatsApp bridge automatic local auth
- 🤖 Provider Expansion & GPT-5 — Added GPT-5 model family support, Xiaomi MiMo, Baidu Qianfan; enhanced Dashscope/ModelArk thinking parameter control, preserves
reasoning_content (chain-of-thought) through full message pipeline
- 💬 Channel Enhancements — Email attachment extraction with MIME filtering, WhatsApp voice transcription, Feishu auto-reaction removal & video downloads, Telegram collapsible tool hints & DM thread support, Langfuse observability integration
- 🔧 Developer Experience — Built-in
grep and glob tools, refactored Tool class with JSON Schema, Python SDK facade for programmatic session isolation, CLI gained --config for multi-instance setups, test suite expanded to 1,142 tests
- 12 messaging platform integrations (Telegram, Discord, WhatsApp, Slack, Feishu, DingTalk, QQ, Email, Matrix, Mochat, Wecom, WeChat)
- MCP (Model Context Protocol) support with SSE transport
- Scheduled tasks and heartbeat jobs
- Local LLM support via vLLM and Ollama
- Docker and Docker Compose deployment
- Clean, readable Python codebase (~4,000 lines)
- Two-stage memory architecture (“Dream” system)
This release focused on dependency cleanup and new platform support:
- 🔧 Removed litellm — Replaced with native SDKs for all providers, reducing dependencies and improving performance
- 🔄 End-to-end streaming — Full streaming support from provider to messaging platform
- 💬 WeChat support — Added WeChat as the 12th messaging platform
- 🔐 Security fix — Patched GHSA-4gmr-2vc8-7qh3 (CVE-2026-33654), a zero-click indirect prompt injection vulnerability in the email channel
- ⚡ Performance improvements — Native SDKs reduce overhead and improve response times
Latest release with 57 PRs merged and 29 new contributors:
- 🔐 Security hardening — Continued improvements to authorization and access control
- 🖥️ Multi-instance enhancements — Better
--config and --workspace handling
- 🔧 MCP improvements — SSE transport, better failure handling
- 🛠️ Tooling enhancements — Auto-casting, validation improvements
- 🌐 Provider support — Azure OpenAI, Alibaba Cloud API support
- ⏰ Cron improvements — External job reload, better context handling
- 💬 Channel updates — Improvements across all platforms:
- Telegram: Proxy fixes, group topics,
/stop command
- Feishu: Rich text parsing, transcription support
- DingTalk: Group chat support
- QQ: Group message handling
- Discord: Group policy, attachments
- WhatsApp: Media support (images, docs, video)
- Matrix: Media normalization
⚠️ Security Note: This release includes all security fixes from v0.1.3.post7 (CVE-2026-2577).
This release focused on real-world polish with 58 PRs merged and 29 new contributors:
- 🔐 Safer by default — Authorization-bypass fixes and stronger
allowFrom handling
- Fixed allowlist bypass via sender_id token splitting (#1677)
- Tighter access control defaults
- 🖥️ Multi-instance support —
--config support, improved runtime path handling
- CLI agent supports
--workspace / --config for multi-instance deployments
- 🔧 MCP improvements — SSE transport support, smarter auto-detection
- Better cancellation/failure handling
- 🛠️ Tooling enhancements — Auto-casting tool params, safer validation
- ReadFileTool size limits to prevent OOM errors
- Better cancellation/failure handling
- 🌐 Provider support — Azure OpenAI, Alibaba Cloud Coding Plan API
- Prompt-caching affinity headers
- ⏰ Cron improvements — External
jobs.json reload, better job context handling
- 💬 Channel updates — Major improvements across all platforms:
- Telegram: Proxy fixes, group topics support,
/stop command, streaming messages
- Feishu: Rich text parsing, table/card splitting, Groq Whisper transcription
- DingTalk: Group chat support
- QQ: Group message handling, markdown sending
- Discord: Group policy, attachments support
- WhatsApp: Media support (images, docs, video)
- Slack: Fallback fixes
- Matrix: Media normalization
⚠️ Security Note: Versions v0.1.3.post4 through v0.1.3.post6 have a critical WhatsApp security vulnerability (CVE-2026-2577). This release includes comprehensive security fixes for GHSA-g6cf-xjwr-pm77 (authorization bypass vulnerability, CVSS 9.0).
Based on development patterns and community feedback:
- More messaging platform integrations
- Enhanced MCP server ecosystem
- Improved resource management
- Better Raspberry Pi optimization
- Maintain minimal codebase philosophy
- Expand tool and plugin ecosystem
- Enhanced multi-agent collaboration
- Enterprise deployment options
- One of the fastest-growing AI agent projects in early 2026
- Featured in AI and developer communities
- Widely adopted in homelab and self-hosting communities
- Recommended for learning AI agent architecture
- 163+ contributors by March 2026
Any questions?
Feel free to contact us. Find all contact information on our contact page.