AgenTool Pieces for Activepieces

Custom Activepieces pieces for integrating with the AgenTool AI Assistant platform.

Features

This package provides 22 actions and 2 triggers for building AI-powered workflows:

Core Assistant Actions

• Chat with Assistant - Send messages and get AI responses
• List Assistants - Get all your AI assistants
• Get Assistant - Get details of a specific assistant
• Create Thread - Start a new conversation
• Get Thread - Retrieve thread messages

Text Analysis Actions

• Classify Text - Categorize text into predefined categories
• Analyze Sentiment - Detect sentiment and emotions
• Summarize Text - Generate concise summaries
• Translate Text - Translate between 20+ languages
• Detect Language - Identify text language
• Extract Entities - Find people, places, organizations, etc.
• Extract Keywords - Pull out important terms
• Detect Intent - Understand user intentions
• Question & Answer - Answer questions from context
• Rewrite Text - Transform text style (professional, casual, etc.)
• Check Guardrails - Validate content against safety rules

AI/ML Actions

• Generate Embeddings - Create vector embeddings for semantic search
• Compute Similarity - Calculate text similarity scores

Image Actions

• Analyze Image - Use AI vision to understand images
• Generate Image - Create images with DALL-E

Audio Actions

• Generate Audio (TTS) - Convert text to speech
• Transcribe Audio - Convert speech to text with Whisper

Triggers

• New Message - Trigger on new assistant messages
• Webhook - Receive AgenTool webhook events

Installation

Option 1: Docker (Recommended)

The pieces are automatically available when running Activepieces with the AgenTool docker-compose:

docker-compose -f docker-compose.activepieces.yml up -d

Option 2: Manual Installation

1. Clone this repository into your Activepieces pieces directory:

cd /path/to/activepieces/packages/pieces/custom
git clone https://github.com/agentool/pieces-agentool agentool

2. Install dependencies:

cd agentool
npm install

3. Build the pieces:

npm run build

4. Add to Activepieces environment:

AP_DEV_PIECES=@activepieces/piece-agentool

Authentication

Embedded Mode (Recommended)

When running Activepieces as part of AgenTool, authentication is automatic via environment variables:

• AGENTOOL_API_URL - AgenTool API endpoint
• AGENTOOL_API_KEY - API key (auto-generated)
• TENANT_NUMBER - Multi-tenant identifier

Standalone Mode

Enter your AgenTool API key when connecting the piece:

1. Go to AgenTool → Settings → API Keys
2. Generate a new key
3. Paste it in the Activepieces connection dialog

Usage Examples

Example 1: Customer Support Classification

Trigger: Webhook (email received)
   ↓
Action: Classify Text
   - Categories: ["billing", "technical", "general", "urgent"]
   ↓
Action: Chat with Assistant
   - Use classification to route to appropriate assistant

Example 2: Content Translation Pipeline

Trigger: New content posted
   ↓
Action: Detect Language
   ↓
Action: Translate Text
   - Target: English, Spanish, French
   ↓
Action: Summarize Text
   - Style: Brief

Example 3: Image Analysis Workflow

Trigger: Image uploaded
   ↓
Action: Analyze Image
   - Get description and labels
   ↓
Action: Generate Embeddings
   - Create searchable vector
   ↓
Action: Chat with Assistant
   - Provide context for customer queries

Development

Building

npm run build

Watching for changes

npm run watch

Project Structure

pieces-agentool/
├── src/
│   ├── index.ts                 # Main piece definition
│   └── lib/
│       ├── common/
│       │   └── client.ts        # AgenTool API client
│       ├── actions/
│       │   ├── chat.ts
│       │   ├── classify.ts
│       │   ├── sentiment.ts
│       │   └── ...
│       └── triggers/
│           ├── new-message.ts
│           └── webhook.ts
├── package.json
├── tsconfig.json
└── README.md

API Reference

All actions use the AgenTool REST API. See the AgenTool API Documentation for full details.

Support

• Documentation: https://agentool.org/docs
• Issues: https://github.com/agentool/pieces-agentool/issues
• Email: [email protected]

License

MIT License - see LICENSE file for details.