Agent700 n8n Custom Nodes

Production-ready custom n8n nodes for integrating with the Agent700 API. These nodes provide seamless authentication, chat interactions, context management, and more.

Table of Contents

• Installation
• Authentication
• Node Documentation
• Workflow Examples
• Best Practices
• Troubleshooting

Installation

Prerequisites

• n8n installed and running
• Node.js 18+ and npm
• Agent700 account credentials

Build the Package

First, build the package from source:

cd Agent700-prod-nodes
npm install
npm run build

This compiles TypeScript to JavaScript in the dist/ folder.

Manual Installation Methods

Choose the installation method that matches your n8n setup:

Option 1: Docker Setup (Recommended for Testing)

If you're using Docker Compose, you have two options:

Option A: Volume Mount (Recommended)

1. Update your docker-compose.yml to mount the built package:

   volumes:
     - ./n8n-data:/home/node/.n8n/data
     - ./Agent700-prod-nodes:/home/node/.n8n/custom/Agent700-prod-nodes

2. Restart your containers:

   docker-compose down
   docker-compose up -d

3. Install dependencies inside the container:

   docker exec <CONTAINER_NAME> sh -c "cd /home/node/.n8n/custom/Agent700-prod-nodes && npm install --production"

Option B: Copy into Container

1. Find your n8n container:

   docker ps | grep n8n

2. Copy the built package into the container:

   docker cp Agent700-prod-nodes <CONTAINER_NAME>:/home/node/.n8n/custom/

3. Install dependencies inside container:

   docker exec <CONTAINER_NAME> sh -c "cd /home/node/.n8n/custom/Agent700-prod-nodes && npm install --production"

4. Restart container:

   docker restart <CONTAINER_NAME>

Option 2: Local n8n Installation (Non-Docker)

If you have n8n installed locally (not in Docker):

1. Copy to n8n custom directory:

   # Find your n8n custom directory (usually ~/.n8n/custom)
   cp -r Agent700-prod-nodes ~/.n8n/custom/
      
   # Install production dependencies
   cd ~/.n8n/custom/Agent700-prod-nodes
   npm install --production

2. Set environment variable (if needed):

   export N8N_CUSTOM_EXTENSIONS=~/.n8n/custom

3. Restart n8n:

   # If running as service
   systemctl restart n8n
      
   # Or if running manually
   n8n start

Option 3: Using npm link (For Development)

For active development with hot reloading:

1. Build and link your package:

   cd Agent700-prod-nodes
   npm install
   npm run build
   npm link

2. Link in n8n directory:

   # If n8n is installed globally
   cd $(npm root -g)/n8n
   npm link agent700-prod-nodes
      
   # Or if n8n is in a specific directory
   cd /path/to/n8n
   npm link agent700-prod-nodes

3. Restart n8n

Note: After making code changes, rebuild (npm run build) and restart n8n for changes to take effect.

Verify Installation

After installation, verify everything works:

1. Check nodes appear in n8n UI:

   • Open n8n interface (typically http://localhost:5678)
   • Create a new workflow
   • Search for "Agent700" - you should see all nodes available:
     • Agent700 Agent
     • Agent700 Context Library

2. Check n8n logs for errors:

   # Docker
   docker logs <CONTAINER_NAME>
      
   # Local
   n8n start --log-level=debug

Troubleshooting Installation

Nodes don't appear:

• Verify dist/ folder exists and contains .node.js files
• Check package.json has correct n8n.nodes array
• Ensure file permissions are correct (Docker: check container user permissions)
• Restart n8n after installation
• Check n8n logs for specific error messages

Dependencies missing:

• Run npm install --production in the custom nodes directory
• For Docker: docker exec <CONTAINER> sh -c "cd /home/node/.n8n/custom/Agent700-prod-nodes && npm install --production"

Build errors:

• Ensure TypeScript is installed: npm install
• Check for TypeScript errors: npm run build
• Verify Node.js version is 18+

Authentication

How Authentication Works

All Agent700 nodes authenticate using an App Password configured directly in the node parameters. Nodes automatically handle authentication on each request - no manual token copying needed!

Setting Up Authentication

1. Get your App Password from the Agent700 web interface

   • Format: app_a7_ followed by 32 characters
   • Example: app_a7_12345678901234567890123456789012

2. Configure in Node Parameters:

   • Base URL: https://api.agent700.ai (default)
   • App Password: Your app password token (required)
   • Nodes automatically use this to obtain access tokens

Authentication Flow

1. Node sends App Password to /api/auth/app-login
2. API returns an access token
3. Node uses Bearer token for all subsequent API calls
4. Token is obtained fresh for each execution

Node Documentation

1. Agent700 Agent

Purpose: Send messages to agents and get structured responses

Key Features:

• Auto-authenticates using App Password
• Agent ID via manual entry
• Simplify output option for cleaner responses
• Full n8n UX guidelines compliance

Parameters:

• Base URL: https://api.agent700.ai (default)
• App Password (required): Your Agent700 app password token
• Resource: Chat (single resource)
• Operation: Send Message
• Agent ID (optional): Enter the Agent UUID manually
• Message (required): Your message to send
• Simplify (default: true): Return simplified output with key fields only

Output (Simplified):

{
  "response": "Agent response",
  "finish_reason": "stop",
  "scrubbed_message": "...",
  "error": null,
  "prompt_tokens": 100,
  "completion_tokens": 50
}

Output (Full): Returns complete API response with all fields.

Example:

1. Add "Agent700 Agent" node
2. Enter App Password
3. Enter Agent ID (UUID from Agent700 web interface)
4. Enter message: "What is AI?"
5. Enable Simplify for cleaner output (optional)
6. Execute

2. Agent700 Context Library

Purpose: Manage alignment data (key-value storage) with encryption at rest

Key Features:

• Full CRUD operations following n8n vocabulary
• Pattern matching and query operations
• JSON construction from patterns
• Auto-authenticates using App Password

Resource: Entry

Operations:

• Get: Retrieve a single entry by key
• Get Many: List all entries
• Create: Create a new entry
• Update: Update an existing entry (with optional key renaming)
• Upsert: Create or update an entry (upsert)
• Delete: Delete an entry (returns { deleted: true, key })
• Query: List key/value pairs matching a pattern
• Query + Construct: Construct JSON from pattern matches using a template

Example (Upsert):

1. Add "Agent700 Context Library" node
2. Enter App Password
3. Resource: Entry
4. Operation: "Upsert"
5. Key: "user_preference"
6. Value: {"theme": "dark_mode"}
7. Execute

Example (Delete):

1. Operation: "Delete"
2. Key: "old_key"
3. Returns: { "deleted": true, "key": "old_key" }

Workflow Examples

Example 1: Simple Chat Workflow

Use Case: One-off questions, simple Q&A

Steps:

1. Manual Trigger → Start workflow
2. Agent700 Agent → Send message
   • Enter App Password
   • Select Agent ID
   • Message: "What is machine learning?"
3. Display Response → Show result

Node Flow:

Manual Trigger → Agent700 Agent → Display Response

When to Use:

• Quick questions
• Single message interactions
• No conversation history needed

Example 2: Chat with Conversation Context

Use Case: Multi-turn conversations, follow-up questions

Note: Conversation context feature is not available in v2. For multi-turn conversations, manually include previous messages in your prompt or use the Context Library to store conversation history.

Steps:

1. Manual Trigger → Start workflow
2. Agent700 Agent → First message
   • Enter App Password
   • Select Agent ID
   • Message: "Explain quantum computing"
3. Agent700 Agent → Follow-up
   • Include previous context in message
   • Message: "Based on your previous explanation, how does it differ from classical computing?"
4. Display Response → Show result

Node Flow:

Manual Trigger → Agent700 Agent → Agent700 Agent (with context) → Display

Example 3: URL Evaluation Workflow

Use Case: Content analysis, privacy policy scanning, URL validation

Steps:

1. Manual Trigger → Start workflow
2. Get URLs → Retrieve URLs (from Context Library or input)
3. Split in Batches → Process one at a time
4. Agent700 Agent → Evaluate each URL
   • Message: "Analyze this URL for privacy concerns: {{$json.url}}"
5. Save Results → Store in Context Library
6. Aggregate → Combine all evaluations

Node Flow:

Trigger → Get URLs → Split → Chat → Save → Aggregate

Advanced Version:

Trigger → Context Library (List) → Loop → Chat → Context Library (Upsert) → Summary

Example 4: Context Library Management Workflow

Use Case: Dynamic context injection, data-driven conversations

Steps:

1. Manual Trigger → Start workflow
2. Agent700 Context Library → List all data
   • Operation: "List All Data"
3. Process Data → Filter/transform as needed
4. Agent700 Context Library → Upsert new data
   • Operation: "Upsert Data"
   • Key: "user_context"
   • Value: "{{$json.processed_data}}"
5. Agent700 Agent → Use context in conversation
   • Message: "Based on this context: {{$json.user_context}}, answer my question"

Node Flow:

Trigger → Context Library (List) → Process → Context Library (Upsert) → Chat

Example 8: Error Handling Workflow

Use Case: Production workflows, reliability-critical applications

Steps:

1. Manual Trigger → Start workflow
2. Agent700 Agent → Attempt chat
3. Error Handler → Catch errors
4. Retry Logic → Retry on failure (with delay)
5. Fallback Response → Use cached/default response if all retries fail

Node Flow:

Trigger → Chat → Error Handler → Retry → Fallback

Implementation Tips:

• Use "Continue on Fail" option in nodes
• Implement retry with exponential backoff
• Cache successful responses for fallback

Example 9: Batch Processing Workflow

Use Case: Bulk operations, data processing pipelines

Steps:

1. Manual Trigger → Start workflow
2. Get Items → Retrieve items to process (from Context Library, database, etc.)
3. Split in Batches → Process in batches
4. Agent700 Agent → Process each item
   • Message: "Process this item: {{$json.item}}"
5. Aggregate Results → Combine all results
6. Save → Store aggregated results

Node Flow:

Trigger → Get Items → Split → Chat (per item) → Aggregate → Save

Performance Tips:

• Process in parallel batches
• Use Continue on Fail for individual items
• Aggregate results efficiently

Best Practices

When to Use Which Node

• Agent Node: Regular chat interactions, sending messages to agents
• Context Library Node: Data storage, context injection, pattern matching

Authentication Management

1. Use App Passwords

   • Required for all nodes
   • Can be revoked individually
   • Better audit trail
   • Format: app_a7_ + 32 characters

2. Store App Passwords Securely

   • Use n8n's parameter encryption
   • Never hard-code in workflows
   • Rotate app passwords regularly
   • Consider using n8n environment variables for sensitive values

3. One App Password Per Environment

   • Separate app passwords for dev/staging/prod
   • Use different Agent IDs per environment

Error Handling

1. Enable Continue on Fail

   • For batch processing
   • When individual failures shouldn't stop workflow

2. Implement Retry Logic

   • For transient errors (network, timeouts)
   • Use exponential backoff

3. Log Errors Properly

   • Use n8n's error handling
   • Store error details in Context Library for debugging

Performance Tips

1. Message Context

   • Include previous messages manually in prompts when needed
   • Use Context Library to store conversation history
   • Limit context size to avoid token limits

2. Batch Processing

   • Process items in parallel when possible
   • Use Split in Batches node
   • Aggregate results efficiently

3. Caching

   • Cache agent configs in Context Library
   • Cache frequently accessed data
   • Use workflow static data for session management

Security

1. SSL/TLS Configuration

   • Use "Strict SSL" in production
   • Only disable for development/testing

2. Token Management

   • Tokens auto-refresh via credentials

3. Data Privacy

   • Be careful with PII in messages
   • Use Context Library encryption features
   • Review scrubbed_message in responses

Troubleshooting

Common Issues

Authentication Fails

Symptoms:

• "Authentication failed" errors
• 401 Unauthorized responses
• "App login did not return accessToken" errors

Solutions:

1. Verify App Password is correct (format: app_a7_ + 32 chars)
2. Check API Base URL is correct (https://api.agent700.ai)
3. Verify App Password is valid in Agent700 web interface
4. Try creating a new app password
5. Check network connectivity
6. Ensure App Password parameter is set in node (not empty)

Node Not Appearing

Symptoms:

• Can't find Agent700 nodes in n8n

Solutions:

1. Verify installation path is correct (see Installation section)
2. Check dist/ folder exists and contains compiled .node.js files
3. Verify package.json n8n.nodes array matches actual file paths
4. Check file permissions (Docker: ensure container user can read files)
5. Restart n8n after installation
6. Check n8n logs for errors: docker logs <CONTAINER> or n8n start --log-level=debug
7. Verify TypeScript compiled successfully (npm run build)
8. For Docker: Ensure volume mount path is correct in docker-compose.yml
9. Install dependencies: npm install --production in the custom nodes directory

Agent ID Not Found

Symptoms:

• "Agent not found" errors

Solutions:

1. Verify the Agent UUID is correct (copy from Agent700 web interface)
2. Verify App Password is correct and has access to the agent
3. Check you have access to agents in Agent700 account
4. Verify API Base URL is correct

API Errors

Symptoms:

• 4xx/5xx HTTP errors
• "API Error" messages

Solutions:

1. Check error details in node output
2. Verify Agent UUID is correct
3. Check API rate limits
4. Review API documentation for endpoint changes
5. Enable Continue on Fail to see detailed errors

SSL/TLS Issues

Symptoms:

• Certificate errors
• Connection refused

Solutions:

1. Use "Strict SSL" in production
2. Check API Base URL uses HTTPS
3. Verify certificate is valid
4. Only use "Allow Self-Signed" for development

Debugging Tips

1. Check Node Output

   • Look at json output for error details
   • Check status field for operation results

2. Enable Continue on Fail

   • See what errors occur
   • Don't stop workflow on first error

3. Use Context Library for Logging

   • Store debug information
   • Track workflow execution

4. Test Individual Nodes

   • Test each node separately
   • Verify credentials work
   • Check API connectivity

Getting Help

1. Check n8n Logs

   • Look for error messages
   • Check execution logs

2. Review API Documentation

   • Agent700 API docs
   • n8n node development docs

3. Test with Simple Workflow

   • Start with basic chat
   • Add complexity gradually

Workflow Templates

Ready-to-use workflow templates are available in the workflows/ folder:

• simple-chat.json - Basic chat workflow
• chat-with-context.json - Conversation context example
• url-evaluation.json - URL evaluation workflow
• context-library-management.json - Context Library operations
• error-handling.json - Error handling example
• batch-processing.json - Batch processing workflow

Note: Workflow templates from v1 may need updates for v2:

• Replace credential references with App Password parameter
• Update node type names (agent700Chat → agent700Agent)
• Update operation names in Context Library node

To import:

1. In n8n, go to Workflows → Import from File
2. Select the JSON file from workflows/ folder
3. Configure App Password in each node
4. Update Agent IDs if needed
5. Execute and customize

License

MIT License - see LICENSE file for details.

Support

• Check n8n documentation for general n8n issues
• Review Agent700 API documentation for API-specific issues
• Create issues in the repository for bugs or feature requests