VoltAgent is an open source TypeScript framework for building AI agents. This article explores real-world examples applications built with VoltAgent, complete with source code and technical details.
What Is VoltAgent?
VoltAgent is a TypeScript-based AI agent framework. It provides memory management, tools, observability, and sub-agent coordination.
WhatsApp Order Agent
A chatbot implementation that accepts food orders through natural conversation on WhatsApp, queries menu items from a database, and maintains conversation history.
Technical Details:
- WhatsApp Business API webhook integration
- SQLite database for menu and order management
- Memory API for conversation history persistence
- Natural language understanding for order parameter extraction
Resources:
YouTube to Blog Agent
A multi-agent system that extracts transcripts from YouTube video URLs and generates Markdown-formatted blog posts.
Technical Details:
- Supervisor agent coordinates 3 sub-agents
- MCP tools for YouTube transcript API integration
- Shared memory for inter-agent data transfer
- Structured Markdown output generation
Resources:
AI Ads Generator Agent
An agent that scrapes landing pages to extract brand information and generates Instagram-formatted visuals using Google Gemini.
Technical Details:
- BrowserBase Stagehand for headless browser control
- Color palette and typography extraction from DOM
- Google Gemini multimodal API for image generation
- 1080x1080 Instagram post format rendering
Resources:
AI Recipe Generator Agent
A recipe recommendation system based on ingredient lists, dietary restrictions, and time parameters.
Technical Details:
- Zod schema validation for ingredients and dietary preferences
- Nutrition database integration
- Structured output for step-by-step instructions
- Portion and time calculation algorithms
Resources:
AI Research Assistant Agent
A research workflow implementation where multiple agents work in parallel for data collection and analysis.
Technical Details:
- 4 different research agents running in parallel
- Type-safe workflow chaining
- OpenTelemetry traces for inter-agent dependency visualization
- Markdown-formatted research report output
Resources:
More Examples
Integration Examples
→ GitHub Repository Analyzer
This example demonstrates how agents can analyze repository code and automatically generate summaries about project structure, dependencies, and potential issues.
→ RAG Chatbot
A document-grounded conversational bot that retrieves relevant information from your knowledge base and provides responses with proper citations.
→ Tavily Web Search
Integrate real-time web search capabilities into your agents, allowing them to augment responses with up-to-date information from the internet.
Vector Database & RAG
→ Chroma Vector Database
This example shows how to implement RAG (Retrieval-Augmented Generation) using Chroma, demonstrating both automatic retrieval and tool-driven retrieval patterns for enhanced context.
→ Pinecone Vector Search
Build semantic search capabilities using Pinecone's vector database, enabling your agents to find contextually similar information through embeddings.
→ Qdrant Vector Database
Compare two different retrieval strategies: retriever-on-every-turn where documents are fetched automatically, versus LLM-decides where the model determines when to search.
→ Postgres with pgvector
Use PostgreSQL with the pgvector extension for both structured data storage and semantic similarity search in a single database.
LLM Providers
→ Anthropic Claude
Connect your agents to Anthropic's Claude models through the AI SDK, giving you access to advanced reasoning and long-context capabilities.
→ Google Gemini AI
Integrate Google's Gemini models into your VoltAgent applications using the AI SDK provider for multimodal AI capabilities.
→ Google Vertex AI
Deploy agents using Google Cloud's Vertex AI platform, leveraging enterprise-grade infrastructure and model management.
→ Groq LPU Inference
Achieve ultra-low latency responses by running your agents on Groq's specialized LPU (Language Processing Unit) hardware.
→ Amazon Bedrock
Configure your agents to use AWS Bedrock's foundation models, accessing a variety of AI models through Amazon's managed service.
→ xAI Grok
Power your agents with xAI's Grok models for real-time understanding and generation capabilities.
MCP (Model Context Protocol)
→ MCP Client Basics
Learn how to connect your agents to Model Context Protocol servers and invoke their tools, enabling standardized integration with external services.
→ Custom MCP Server
Build your own MCP server that exposes custom tools to agents, allowing you to create reusable tool ecosystems across different agent applications.
→ Composio MCP Integration
Integrate Composio's suite of third-party application actions into your agents through the Model Context Protocol interface.
→ Google Drive MCP
Enable your agents to browse folders and read files from Google Drive using an MCP server connection.
→ Hugging Face MCP
Access HuggingFace's vast collection of models and tools through MCP, allowing your agents to leverage specialized AI capabilities.
→ Zapier MCP Integration
Connect your agents to thousands of applications through Zapier's automation platform using MCP integration.
→ Peaka MCP Integration
Integrate Peaka's data federation and query services into your agents through MCP tools for unified data access.
Deployment Platforms
→ Next.js Integration
Build a React-based frontend that communicates with VoltAgent APIs, featuring streaming responses for real-time agent interactions.
→ Nuxt Integration
Create a Vue/Nuxt application that seamlessly integrates with VoltAgent's backend services for server-side rendered agent experiences.
→ Cloudflare Workers Deployment
Deploy your agents to Cloudflare's edge network using the Hono server adapter for global, low-latency serverless execution.
→ Netlify Functions Deployment
Host your agent APIs as serverless functions on Netlify, enabling easy deployment with automatic scaling and CDN distribution.
Advanced Patterns
→ Supervisor and Sub-agents
Implement hierarchical agent systems where a supervisor agent coordinates multiple specialized sub-agents, each handling specific aspects of complex tasks.
→ Multi-step Workflows
Create sophisticated multi-step workflows using createWorkflowChain, including human-in-the-loop approval steps for critical decisions.
→ Working Memory Management
Maintain per-conversation facts and context that persist across interactions, with built-in tools for reading and updating stored information.
→ Semantic Vector Search
Enable agents to automatically recall relevant context from past conversations using semantic memory and vector similarity search.
→ Client-side Tool Execution
Execute type-safe tools directly in the browser while maintaining security, with a Next.js frontend managing client-side interactions.
→ Agent-to-Agent Communication
Set up HTTP endpoints that allow different agents to communicate with each other, enabling distributed multi-agent architectures.
Tools & Utilities
→ Zod-typed Tool Creation
Learn how to create type-safe tools using Zod schemas, with support for cancellation, streaming responses, and full TypeScript inference.
→ Structured Thinking Tool
Give your agents a dedicated thinking tool that enables structured reasoning and step-by-step problem solving before providing final answers.
→ Playwright Browser Automation
Equip your agents with browser automation capabilities using Playwright, enabling them to interact with web pages, fill forms, and extract data.
→ Dynamic Prompt Generation
Build prompts programmatically from templates and runtime data, allowing you to customize agent behavior based on context and user input.
→ Dynamic Parameter Validation
Validate and inject runtime parameters into your agents using Zod schemas, ensuring type safety and proper input handling at execution time.
Observability & Evaluation
→ OpenTelemetry Trace Example
Set up OpenTelemetry tracing with VoltOps integration, allowing you to inspect detailed execution spans and understand your agent's decision-making process.
→ Langfuse Integration
Export traces and metrics to Langfuse's observability platform for comprehensive monitoring, debugging, and performance analysis of your AI agents.
→ Live Agent Evaluations
Run real-time evaluations on your agents during development, helping you catch issues and validate behavior changes immediately.
→ Offline Batch Evaluations
Test your agents against predefined datasets in batch mode, enabling systematic regression testing and quality assurance.
→ ViteVal Evaluation
Evaluate agent performance and prompt effectiveness using ViteVal's testing framework for systematic quality measurement.
→ Telemetry Exporter
Configure custom telemetry exports to send traces, metrics, and logs to external observability platforms like Datadog or New Relic.
Voice & Audio
→ OpenAI Text-to-Speech
Convert your agent's text responses into natural-sounding speech using OpenAI's TTS API with multiple voice options.
→ ElevenLabs Voice Generation
Generate high-quality, realistic voice audio from agent responses using ElevenLabs' advanced text-to-speech technology.
→ xAI Voice Synthesis
Integrate xAI's audio models to synthesize voice output from your agent's text responses with natural prosody.
Security & Storage
→ JWT Authentication
Secure your agent endpoints with JWT token verification, ensuring only authorized users can access your AI services.
→ Supabase Integration
Leverage Supabase for user authentication and database operations within your agent tools, combining auth and data storage in one platform.
→ Turso Database
Persist agent memory and conversation history using Turso's distributed LibSQL database for fast, edge-optimized storage.
→ VoltOps Managed Memory
Use VoltOps' managed memory service through a REST adapter, offloading memory management for production-scale agent deployments.
Core Examples
→ Minimal Starter Project
Get started with the simplest possible VoltAgent setup featuring a single agent and local development server.
→ Output Guardrails
Add validation rules and schema enforcement to your agent outputs, ensuring responses always conform to your required format.
→ Lifecycle Hooks
Implement lifecycle hooks to add logging, authentication, or custom middleware at different stages of agent execution.
→ Custom REST Endpoints
Extend your VoltAgent server with custom REST routes for additional functionality beyond standard agent endpoints.
→ Retriever API
Explore the basics of VoltAgent's retriever API for fetching relevant context to augment agent responses.
→ Vercel AI SDK
Integrate VoltAgent with Vercel's AI SDK for streaming responses and seamless deployment on Vercel's platform.
Getting Started
Create a new VoltAgent project:
npm create voltagent-app@latest
This scaffolds a TypeScript project with agent definitions, workflow examples, and VoltOps integration configured.
VoltAgent is open source. Contributions are welcome on GitHub.