Skip to main content

​
Prerequisites

Before building your first agent, ensure you have:

​
Quick Start

​
1. Clone and Setup

# Clone the repository
git clone https://github.com/adenhq/hive.git
cd hive

# Run Python environment setup
./scripts/setup-python.sh
This installs:
  • framework - Core agent runtime and graph executor
  • aden_tools - MCP tools for agent capabilities
  • All required dependencies

​
2. Install Claude Code Skills

# One-time installation of building skills
./quickstart.sh

​
3. Build Your Agent

Launch Claude Code and use the building skill: 
claude> /building-agents
The Coding Agent will guide you through:
  1. Defining your goal - What should the agent accomplish?
  2. Setting success criteria - How do you measure success?
  3. Adding nodes - LLM, Router, Function, or Human nodes
  4. Connecting edges - Define the execution flow
  5. Testing - Validate the agent behavior

​
Example: Building a FAQ Agent

Let’s build a simple agent that answers frequently asked questions.

​
Define the Goal

Goal: Answer customer FAQs about our product
Success: Correctly answer 90% of common questions
Constraints: Escalate to human if confidence < 0.7

​
Generated Agent Structure

exports/faq_agent/
├── __init__.py
├── agent.json          # Agent graph specification
├── tools.py            # Custom tool implementations
├── prompts/
│   └── answer.txt      # Prompt templates
└── tests/
    └── test_agent.py   # Test cases

​
agent.json

{
  "name": "faq-agent",
  "version": "1.0.0",
  "goal": "Answer customer FAQs accurately",
  "nodes": [
    {
      "id": "classify",
      "type": "llm",
      "model": "claude-sonnet-4-5-20250929",
      "prompt": "Classify this question into a category: {{question}}"
    },
    {
      "id": "answer",
      "type": "llm",
      "model": "claude-sonnet-4-5-20250929",
      "prompt": "Answer this FAQ: {{question}}\nCategory: {{category}}"
    },
    {
      "id": "confidence_check",
      "type": "router",
      "conditions": [
        {"path": "respond", "when": "confidence >= 0.7"},
        {"path": "escalate", "when": "confidence < 0.7"}
      ]
    },
    {
      "id": "escalate",
      "type": "human",
      "prompt": "Please answer: {{question}}",
      "timeout": "1h"
    }
  ],
  "edges": [
    {"from": "classify", "to": "answer", "type": "on_success"},
    {"from": "answer", "to": "confidence_check", "type": "on_success"},
    {"from": "confidence_check", "to": "respond", "when": "confidence >= 0.7"},
    {"from": "confidence_check", "to": "escalate", "when": "confidence < 0.7"}
  ]
}

​
4. Test Your Agent

claude> /testing-agent
Or run manually: 
PYTHONPATH=core:exports python -m faq_agent run --input '{"question": "What is your refund policy?"}'

​
5. Run Your Agent

PYTHONPATH=core:exports python -m faq_agent run --input '{
  "question": "How do I reset my password?"
}'
Output: 
{
  "answer": "To reset your password, click 'Forgot Password' on the login page...",
  "confidence": 0.95,
  "category": "account_management"
}

​
Project Structure

When you build agents with Aden, they’re exported to the exports/ directory: 
hive/
├── core/               # Framework runtime
├── tools/              # MCP tools package
├── exports/            # Your agents live here
│   ├── faq_agent/
│   ├── support_agent/
│   └── sales_agent/
└── .claude/            # Claude Code skills

​
Next Steps

Node Types

Learn about LLM, Router, Function, and Human nodes

Edge Configuration

Configure success, failure, and conditional edges

Human-in-the-Loop

Add human intervention points

Testing Agents

Write and run agent tests