What is HNO?
HNO is a multi-agent system framework built with Go. It uses Go's concurrency model, static typing, deployment model, and standard tooling; workload-specific performance claims are documented only when a reproducible benchmark is available.
Key Features
🚀 Measured Performance
- Agent-construction benchmark results and the exact environment are recorded in Performance.
- The benchmark uses a local
MockModeland measures framework allocation, not LLM latency or service throughput. - No Go-vs-Python speedup or memory ratio is published without an apples-to-apples benchmark.
- Native Concurrency: Go goroutines are available for application-level concurrency.
🤖 AgentOS HTTP Server
HNO includes AgentOS, an HTTP server with:
- RESTful API with OpenAPI 3.0 specification
- Session management for multi-turn conversations
- Thread-safe agent registry
- Health monitoring and structured logging
- CORS support and request timeout handling
🧩 Flexible Architecture
Three core abstractions for different use cases:
- Agent - Autonomous AI agents with tool support and memory
- Team - Multi-agent collaboration with 4 coordination modes
- Sequential, Parallel, Leader-Follower, Consensus
- Workflow - Step-based orchestration with 5 primitives
- Step, Condition, Loop, Parallel, Router
🔌 Multi-Model Support
Built-in support for multiple LLM providers:
- OpenAI - GPT-4, GPT-3.5 Turbo, etc.
- Anthropic - Claude 3.5 Sonnet, Claude 3 Opus/Sonnet/Haiku
- Ollama - Local models (Llama 3, Mistral, CodeLlama, etc.)
- DeepSeek - DeepSeek-V2, DeepSeek-Coder
- Google Gemini - Gemini Pro, Flash
- ModelScope - Qwen, Yi models
🔧 Extensible Tools
Following the KISS principle, we provide essential tools with high quality:
- Calculator - Basic math operations
- HTTP - Make HTTP GET/POST requests
- File Operations - Read, write, list, delete with security controls
- Search - DuckDuckGo web search
Easy to create custom tools - see Tools Guide.
💾 RAG & Knowledge
Build intelligent agents with knowledge bases:
- ChromaDB - Vector database integration
- OpenAI Embeddings - text-embedding-3-small/large support
- Automatic embedding generation and semantic search
See RAG Demo for a complete example.
Design Philosophy
KISS Principle
Keep It Simple, Stupid - Focus on quality over quantity:
- A small, inspectable core
- Essential tools
- Pluggable storage integrations
This focused approach aims for:
- Better code quality
- Easier maintenance
- Deployable server features; production suitability depends on the workload
Go Advantages
Why build multi-agent systems with Go?
- Performance - Compiled language, fast execution
- Concurrency - Native goroutines, no GIL
- Type Safety - Catch errors at compile time
- Single Binary - Easy deployment, no runtime dependencies
- Great Tooling - Built-in testing, profiling, race detection
Use Cases
HNO is perfect for:
- Production AI Applications - Deploy with AgentOS HTTP server
- Multi-Agent Systems - Coordinate multiple AI agents
- Application Workflows - Compose multi-step agent tasks
- Local AI Development - Use Ollama for privacy-focused applications
- RAG Applications - Build knowledge-based AI assistants
Quality Metrics
- Test status: Run
go test ./...for the current result - Benchmark status: See the reproducible snapshot in Performance
- Documentation: Guides, API reference, and examples are built with VitePress
- Deployment: Docker and deployment material are provided; production suitability depends on the deployment workload
Next Steps
Ready to get started?
- Quick Start - Build your first agent in 5 minutes
- Installation - Detailed setup instructions
- Core Concepts - Learn about Agent, Team, Workflow
Quick Links
- Embeddings: OpenAI/VLLM usage
- Vector Indexing: Chroma + Redis (optional) + CLI
Community
- GitHub: rexleimo/HNO
- Issues: Report bugs
- Discussions: Ask questions
License
HNO is released under the MIT License.
Inspired by the Agno Python project. HNO is the current name of this Go project; the repository does not define an official expansion of the name.