AgentVoy

AgentVoy is a platform for developing, deploying, and managing AI agent applications that supports both single-agent and multi-agent pipelines. It is designed for users who want to build agentic apps using a variety of popular frameworks and models, providing a unified workflow from project creation to deployment. The tool emphasizes rapid project setup, real-time development insights, and consistent security guardrails across different agent frameworks.

Developers can use AgentVoy’s command-line interface to create new projects with a single command, selecting from seven supported frameworks, including OpenAI Agents SDK, Google ADK, CrewAI, LangGraph, Anthropic SDK, LlamaIndex, and AutoGen—all using Python. For model integration, the platform accommodates a wide range of providers such as OpenAI, Anthropic, Google, Groq, Mistral, and also allows fully local model execution with Ollama. The Streamlit-based chat UI auto-detects available models from environment variables, enabling dynamic model switching without restarting the application. Multi-agent pipelines are supported, allowing sequential workflows where each agent builds upon the previous stage, with real-time visualization of pipeline stages.

A central feature of AgentVoy is its real-time DevTools dashboard, accessible via a web interface. This dashboard provides live tracing of agent operations, including LLM calls, tool invocations, guard checks, and pipeline stages, with trace events streamed over WebSocket. Projects include a FastAPI server, Streamlit chat UI, and cloud configuration files, facilitating both local experimentation and full-scale deployment. Deployment is streamlined with one-command options to Docker, Fly.io, Railway, Google Cloud Platform (Cloud Run), and AWS Lambda, with automatic handling of secrets, configuration, and health checks. The platform can also add deployment capabilities to existing agent projects without altering core agent logic.

Security and operational guardrails are built into every AgentVoy project via a universal agent.guard.yml configuration file. This file enforces permissions, cost limits, iteration caps, prompt injection blocking, and PII detection at runtime, and its settings are automatically integrated into deployment infrastructure. No installation is required to get started; projects can be created and deployed directly using npx commands. AgentVoy is intended for developers seeking a unified, secure, and efficient approach to building and shipping AI agent applications across multiple frameworks and deployment targets.