n8nCopilot

AI agent for n8n workflow automation

An AI-powered CLI assistant that helps you build n8n workflows through natural language. Describe what you want to automate, and it generates, validates, and can deploy workflows directly to your n8n instance.

Project Structure

n8nCopilot/
├── src/
│   ├── config/          # Model configuration
│   ├── tools/           # Agent tools (todos, memory, workflow, search)
│   ├── prompt/          # System prompts
│   ├── ui/              # CLI UI components
│   ├── memory/          # 3-layer memory system
│   └── index.ts         # Entry point
├── .n8ncopilot/          # Runtime memory storage
└── package.json

---

Install

# Clone 
npm install

# Copy and configure env
cp .env.example .env
# Edit .env with your values

# Run
npm start
# or: bun index.ts
# or: npx tsx index.ts

---

Model Support

n8nCopilot works with any LangChain-compatible model:

Provider	Setup
Ollama (local, free)	ollama serve then select model like qwen2.5:14b
OpenAI	Set OPENAI_API_KEY in .env
Anthropic	Set ANTHROPIC_API_KEY in .env
Groq	Set GROQ_API_KEY in .env

Switch model anytime: type /model in the chat.

Recommended models:

• qwen2.5:14b (Ollama) — best for JSON generation locally
• gpt-4o (OpenAI) — most reliable tool calling
• claude-sonnet-4-5-20251022 (Anthropic) — best planning
• llama-3.3-70b-versatile (Groq) — fastest open model

---

Usage

  you › Build a webhook that receives Stripe payments and saves to Postgres with Slack alerts on failure

  ◆  Planning...
  ⟳ write_todos({"tasks":["Search Stripe + Postgres + Slack nodes","Generate workflow","Validate","Add error handling"]})
  ✓ write_todos: Plan created (4 steps)

  ⟳ generate_workflow({"description":"Stripe webhook → Postgres → Slack alerts","trigger":"webhook",...})
  ✓ generate_workflow: Generated "Stripe Payment Processor" — 8 nodes

  ⟳ validate_workflow({...})
  ✓ validate_workflow: ✅ 0 errors · 1 warning · 8 nodes

  Generated workflow: Stripe Payment Processor
  • 8 nodes  •  webhook trigger  •  Postgres + Slack
  • Full error handling with Slack notifications
  • continueOnFail on all external nodes

  Copy the JSON below or type "deploy this" to push to your n8n instance.

---

Commands

Command	Description
/model	Switch model provider
/memory	View long-term memory state
/todos	View current plan
/clear	Clear screen
/help	Show help
/exit	Quit

---

Architecture

n8nCopilot
├── DeepAgent Core (LangGraph createReactAgent)
│   ├── Planning: write_todos / read_todos / update_todo
│   ├── Memory: save_memory / read_memory / list_memory
│   └── Context: set_project_context
├── n8n Knowledge
│   ├── search_nodes (40+ nodes catalog)
│   └── generate_workflow (sub-call with n8n expert prompt)
├── Workflow Pipeline
│   ├── generate_workflow → validate_workflow → add_error_handling
│   └── deploy_workflow / list_workflows (n8n API)
├── 3-Layer Memory System
│   ├── Layer 1: .n8ncopilot/memory/index.json (always loaded)
│   ├── Layer 2: .n8ncopilot/memory/topics/*.json (on-demand)
│   └── Layer 3: .n8ncopilot/memory/feedback.json (grep only)
├── Streaming (dim tool call display)
└── Feedback Learning (thumbs up/down → memory)

---

n8n API Setup

To enable direct deployment:

1. Open n8n → Settings → API
2. Create an API key
3. Add to .env:

   N8N_API_URL=http://localhost:5678
   N8N_API_KEY=your_key_here
   

---

Author

Rahees Ahmed