Skip to content

kadirpekel/hector

Repository files navigation

██╗  ██╗███████╗ ██████╗████████╗ ██████╗ ██████╗ 
██║  ██║██╔════╝██╔════╝╚══██╔══╝██╔═══██╗██╔══██╗
███████║█████╗  ██║        ██║   ██║   ██║██████╔╝
██╔══██║██╔══╝  ██║        ██║   ██║   ██║██╔══██╗
██║  ██║███████╗╚██████╗   ██║   ╚██████╔╝██║  ██║
╚═╝  ╚═╝╚══════╝ ╚═════╝   ╚═╝    ╚═════╝ ╚═╝  ╚═╝

Go Version License A2A Protocol Documentation Go Report Card

Build AI agents without code

A declarative A2A native AI agent platform. Define sophisticated agents through simple YAML configuration.

Built with Go for production performance, single-binary deployment, and true portability.

⚡️ From idea to production agent in minutes, not months.

Getting StartedDocumentationExamples


⚠️ Alpha Version: This project is in active development. Features may change as we refine the platform.


Quick Start

# Install (single binary, no dependencies)
go install github.com/kadirpekel/hector/cmd/hector@latest

# Set API key
export OPENAI_API_KEY="sk-..."

# Run your first agent (zero-config mode)
hector call "What is the capital of France?"

Why Go? Unlike Python-based frameworks, Hector compiles to a single binary with no runtime dependencies. Deploy anywhere—from edge devices to Kubernetes—with consistent performance and minimal resource usage.

Configuration Example

agents:
  assistant:
    llm: gpt-4o
    tools: [search, write_file, execute_command]
    reasoning:
      engine: chain-of-thought
      max_iterations: 100
    memory:
      working:
        strategy: summary_buffer

That's it. No code required.


Why Hector?

For Developers

  • Zero-code development - YAML configuration only
  • Instant setup - Working agent in 5 minutes
  • Advanced memory - Working & long-term memory strategies
  • RAG & semantic search - Built-in vector store integration
  • Rich tool ecosystem - Built-in tools, MCP, and plugins

For Enterprises

  • True interoperability - Native A2A protocol support
  • Multi-agent orchestration - Coordinate specialized agents
  • Production security - JWT auth, API keys, agent-level security
  • Distributed architecture - Local, server, or client modes
  • Multi-transport APIs - REST, SSE, WebSocket, gRPC

For Teams

  • Simple configuration - Human-readable YAML
  • Declarative approach - No code to maintain
  • Built with Go - Production performance, single binary, no dependencies
  • Flexible deployment - Docker, Kubernetes, systemd
  • Extensible platform - Custom plugins via gRPC
  • Open source - AGPL-3.0 licensed

Key Features

Agent Development

  • Multiple LLM providers (OpenAI, Anthropic, Gemini)
  • Slot-based prompt system
  • Structured output (JSON, XML, Enum)
  • Working & long-term memory
  • Two reasoning strategies
  • Session management

Production Features

  • Built with Go - High performance, single binary
  • A2A protocol compliant
  • Multi-agent orchestration
  • Built-in & MCP tools
  • RAG with vector databases
  • JWT authentication
  • Real-time streaming
  • Plugin system (gRPC)

Documentation

📚 Complete Documentation at gohector.dev

Quick Links


Architecture

Hector's layered architecture:

Application (Your Agents)
         ↓
   Hector Runtime
         ↓
    A2A Protocol
         ↓
Multi-Transport Layer (gRPC, REST, JSON-RPC)
         ↓
  Agent Orchestration
         ↓
Core Services (LLM, Memory, Tools, RAG, Tasks)

For detailed architecture, see the Architecture Reference.


Deployment Modes

Three ways to run Hector:

# 1. Local Mode - In-process execution
hector call assistant "Hello"

# 2. Server Mode - Host agents for multiple clients
hector serve --config agents.yaml

# 3. Client Mode - Connect to remote agents
hector call assistant "Hello" --server https://remote:8080

Multi-Agent Example

Coordinate specialized agents:

agents:
  coordinator:
    llm: gpt-4o
    reasoning:
      engine: supervisor
    tools:
      - agent_call
    sub_agents:
      - researcher
      - analyst
      - writer
  
  researcher:
    llm: gpt-4o
    tools: [search]
  
  analyst:
    llm: claude-sonnet-4
    tools: [search]
  
  writer:
    llm: gpt-4o
    tools: [write_file]

See Multi-Agent Tutorial for complete example.


A2A Protocol

Hector is 100% compliant with the Agent-to-Agent protocol, enabling true interoperability:

  • Connect to any A2A-compliant agent
  • Expose your agents for others to use
  • Federation across organizations
  • Standards-based communication

Learn more: A2A Protocol Reference


Examples

Check out configs/ for complete examples:


Contributing

We welcome contributions! Hector is in active development and we're building this in the open.

Ways to contribute:

  • 🐛 Report bugs and issues
  • 💡 Suggest features and improvements
  • 📖 Improve documentation
  • 🔧 Submit pull requests
  • ⭐ Star the repo and spread the word

See Contributing Guidelines for details.


Community


License

AGPL-3.0 - See LICENSE.md for details.

Hector is free and open-source software. You can use, modify, and distribute it under the terms of the AGPL-3.0 license.

For commercial licensing options, please contact the maintainers.


Built with ❤️ by the Hector community

Star on GitHub

Packages

No packages published

Languages