n8n-nodes-berget-ai-agent

n8n node for Berget AI Agent with tool calling capabilities - similar to OpenAI Agent node.

Installation

Community Nodes (Recommended)

1. Open n8n
2. Go to Settings > Community Nodes
3. Click Install a community node
4. Enter: @bergetai/n8n-nodes-berget-ai-agent
5. Click Install

Manual Installation

# In your n8n project
npm install @bergetai/n8n-nodes-berget-ai-agent

Local Development

# Clone this repo
git clone <repo-url>
cd n8n-nodes-berget-ai-agent

# Install dependencies
npm install

# Build project
npm run build

# Link locally for development
npm link
cd /path/to/your/n8n/project
npm link @bergetai/n8n-nodes-berget-ai-agent

Configuration

1. Add the node to your workflow
2. Configure API settings:
   • API Key: Your Berget AI API key
   • Base URL: https://api.berget.ai/v1 (default)
   • Model: Choose from available models
   • System Prompt: Define the agent's behavior
   • User Message: The task or question for the agent
   • Tools: Define tools the agent can use

Available Models

• Llama 3.1 8B Instruct
• Llama 3.3 70B Instruct (recommended for agents)
• GLM-4.6
• DeepSeek-OCR (vision model for image analysis)
• Mistral Small 3.1 24B Instruct 2503
• Qwen3 32B
• GPT-OSS-120B

Features

• ✅ Tool Calling: Define custom tools with JSON Schema
• ✅ Multi-turn Conversations: Agent can make multiple tool calls
• ✅ Flexible Tool Choice: Auto, required, or none
• ✅ Iteration Control: Set maximum number of agent iterations
• ✅ System Prompts: Customize agent behavior
• ✅ Temperature Control: Adjust response randomness
• ✅ Token Limits: Control response length

How It Works

The agent node follows this workflow:

1. Initial Request: User provides a task/question and available tools
2. Agent Processing: Model analyzes the task and decides if tools are needed
3. Tool Detection: If needed, the agent identifies required tool calls with parameters
4. Output: The node outputs tool calls for external execution (does not execute tools itself)
5. Integration: Connect to other n8n nodes to execute tools and feed results back

Important: This node does NOT execute tools automatically. It only detects when tools should be called and outputs the tool call information. You must implement tool execution in your n8n workflow using other nodes.

Tool Definition

Tools are defined using JSON Schema format:

{
  "type": "object",
  "properties": {
    "location": {
      "type": "string",
      "description": "The city and country"
    },
    "units": {
      "type": "string",
      "enum": ["celsius", "fahrenheit"],
      "description": "Temperature units"
    }
  },
  "required": ["location"]
}

Examples

Basic Agent with Weather Tool

{
  "operation": "agent",
  "model": "meta-llama/Llama-3.3-70B-Instruct",
  "systemPrompt": "You are a helpful assistant with access to weather data.",
  "userMessage": "What's the weather in Stockholm?",
  "tools": [
    {
      "name": "get_weather",
      "description": "Get current weather for a location",
      "parameters": {
        "type": "object",
        "properties": {
          "location": {
            "type": "string",
            "description": "City and country"
          }
        },
        "required": ["location"]
      }
    }
  ],
  "options": {
    "temperature": 0.7,
    "max_iterations": 5,
    "tool_choice": "auto"
  }
}

Agent with Multiple Tools

{
  "operation": "agent",
  "model": "meta-llama/Llama-3.3-70B-Instruct",
  "systemPrompt": "You are a research assistant.",
  "userMessage": "Find information about AI and save it.",
  "tools": [
    {
      "name": "search_web",
      "description": "Search the web for information",
      "parameters": {
        "type": "object",
        "properties": {
          "query": { "type": "string" }
        },
        "required": ["query"]
      }
    },
    {
      "name": "save_document",
      "description": "Save text to a document",
      "parameters": {
        "type": "object",
        "properties": {
          "content": { "type": "string" },
          "filename": { "type": "string" }
        },
        "required": ["content", "filename"]
      }
    }
  ]
}

Output Format

The agent node returns:

{
  "response": "Final agent response text or null if tools were requested",
  "tool_calls": [
    {
      "id": "call_abc123",
      "name": "get_weather",
      "arguments": "{\"location\": \"Stockholm, Sweden\"}",
      "iteration": 1
    }
  ],
  "iterations": 2,
  "messages": [...],
  "model": "meta-llama/Llama-3.3-70B-Instruct"
}

Note: If the AI requests tool calls, the node will throw an error with details about the requested tools. This is intentional - you must handle tool execution in your workflow and then continue the conversation.

Testing

Quick Test

# Test node structure
npm test

# Test with real API
BERGET_API_KEY=your-key npm test

# Link locally for n8n testing
npm run test:local

Comparison with OpenAI Agent

This node provides similar functionality to OpenAI's Agent node:

Pricing

See current pricing at berget.ai/models

All prices are in EUR per million tokens.

Support

• 📧 Support: kontakt@berget.ai
• 📖 API Documentation: api.berget.ai/docs
• 🐛 Issues: Report issues in the repository

License

MIT License - See LICENSE file for details.