@grigoreo-fox/n8n-nodes-alt-ai-agent

This is an n8n community node that provides an alternative AI Agent with enhanced memory capabilities. It functions almost identically to the stock n8n AI Agent but adds tool call memory persistence - automatically saving tool interactions to the conversation memory for better context retention across agent iterations.

The core differentiating feature is implemented in toolAwareMemory.ts, which intelligently intercepts and folds tool call summaries into the agent's memory, providing richer context for multi-step reasoning tasks.

n8n is a fair-code licensed workflow automation platform.

Table of Contents

• Installation
• Key Features
• Operations
• Configuration
• Tool Memory Feature
• Limitations
• Compatibility
• Usage Examples
• Resources
• Version History

Installation

Follow the installation guide in the n8n community nodes documentation.

Install via npm:

npm install @grigoreo-fox/n8n-nodes-alt-ai-agent

Or install directly in n8n using Community Nodes:

1. Go to Settings > Community Nodes
2. Enter: @grigoreo-fox/n8n-nodes-alt-ai-agent
3. Click Install

Key Features

🧠 Enhanced Memory with Tool Call Persistence

Unlike the standard AI Agent, this node automatically saves tool call details to memory, including:

• Tool names and arguments used
• Tool execution results and observations
• Formatted summaries for better context retention

🔧 Full Tool Integration

• Connect multiple tools via n8n's AI Tool connection type
• Automatic tool discovery and binding
• Support for custom tool implementations

💬 Flexible Conversation Management

• System message customization
• Memory integration for conversation history
• Configurable iteration limits

Operations

The Alt AI Agent supports the following operations:

• Chat & Execute: Process user messages and execute multi-step plans using connected tools
• Tool Orchestration: Coordinate multiple tool calls within a single agent session
• Memory Integration: Persist conversation context including tool interactions

Configuration

Required Connections

• Chat Model: Connect an AI Language Model (OpenAI, Anthropic, etc.)

Optional Connections

• Memory: Connect a Memory node to persist conversation history
• Tools: Connect one or more Tool nodes for agent capabilities

Node Options

| Option | Type | Default | Description | |--------|------|---------|-------------| | System Message | string | (default prompt) | Initial instructions sent to the agent | | Prompt (User Message) | string | ={{ $json.input }} | The user message/task for the agent | | Max Iterations | number | 10 | Maximum agent reasoning iterations | | Save Tool Calls to Memory | boolean | true | 🔑 Key Feature: Persist tool calls in memory | | Return Intermediate Steps | boolean | true | Include agent reasoning steps in output |

Tool Memory Feature

The core innovation of this node is the toolAwareMemory system (implemented in toolAwareMemory.ts):

How It Works

1. Intercepts Memory Saves: Wraps the standard LangChain memory with a proxy
2. Captures Tool Calls: Extracts tool names, arguments, and results from agent steps
3. Formats Summaries: Creates human-readable tool execution summaries
4. Persists Context: Automatically saves tool call history alongside conversation

Example Tool Memory Output

tool call: web_search({"query": "n8n community nodes"}) => {"results": "Found 10 results about n8n community nodes..."}

tool call: send_email({"to": "[email protected]", "subject": "Report"}) => {"status": "sent", "message_id": "abc123"}

User: Thanks for sending that email! Can you search for more information about workflow automation?

Benefits

• Enhanced Context: Agent remembers what tools were used and their results
• Better Decision Making: Avoid redundant tool calls by referencing previous executions
• Debugging: Clear visibility into tool interaction history
• Continuity: Maintain tool context across conversation turns

Limitations

⚠️ Important Limitations:

• No Structured Output: This node does not support structured output formatting
• No Streaming: Real-time streaming responses are not available
• Memory Dependency: Tool call persistence requires a connected Memory node

These limitations are trade-offs for the enhanced memory functionality.

Compatibility

• Minimum n8n version: 1.0.0
• Tested with: n8n 1.x.x series
• LangChain: Compatible with LangChain 0.3.x
• Node.js: Requires Node.js 18+

Usage Examples

Basic Agent with Tool Memory

1. Connect OpenAI Chat Model
2. Connect Buffer Memory  
3. Connect HTTP Request Tool
4. Enable "Save Tool Calls to Memory" (default: true)
5. Set prompt: "Help me analyze the website example.com"

The agent will automatically save HTTP request details to memory for future reference.

Multi-Step Task with Memory

Prompt: "Search for the latest n8n updates and email me a summary"

Agent Flow:
1. Uses web search tool → Saves search results to memory
2. Analyzes findings → References previous search in context  
3. Uses email tool → Saves email status to memory
4. Provides confirmation → Full tool history available

Resources

• n8n Community Nodes Documentation
• LangChain Agents Documentation
• n8n AI Agent Documentation
• Source Code

Version History

v0.1.0

• Initial release
• Implemented toolAwareMemory system for persistent tool call context
• Full compatibility with n8n AI Tool ecosystem
• Enhanced memory integration with tool call summaries

Contributing

This is an open source project. Contributions, issues, and feature requests are welcome!

License

MIT License - see LICENSE file for details.

Built with ❤️ for the n8n community