Waldiez.io provides open-source tools for building and orchestrating production AI agent systems, combining a visual multi-agent workflow builder with a runtime actor-model orchestration platform. The platform enables engineers to design workflows through a drag-and-drop interface powered by the AG2 framework, while the Wactorz runtime allows agents to spawn dynamically over MQTT and persist across restarts. Supporting deployment from edge devices to cloud environments, it integrates with JupyterLab, VS Code, and a community hub for sharing workflows.
Funding
Funding not disclosed
Founders
Product
Problem
Engineers building production AI systems face significant complexity in orchestrating multi-agent workflows, often requiring extensive coding for agent coordination, state management, and cross-environment deployment. Traditional approaches couple agent definitions to compile-time, making runtime flexibility and crash recovery difficult to achieve without substantial engineering effort.
Solution
Waldiez.io offers two complementary open-source platforms: Wactorz, a runtime actor-model orchestration system, and Waldiez, a visual multi-agent workflow builder. Wactorz lets engineers describe agents in natural language, which then spawn at runtime over MQTT, persist across restarts, and scale from laptop to edge without code changes. Waldiez provides a drag-and-drop visual builder powered by the AG2 framework, supporting multi-LLM integration across major providers and real-time agent collaboration. The platform includes a Community Hub for sharing and forking workflows, with extensions for JupyterLab, VS Code, and a dedicated Studio environment.
Target Audience
Primary customers are AI engineers and DevOps teams building production multi-agent systems, as well as data scientists and researchers using JupyterLab or VS Code who need visual workflow design and orchestration capabilities.
Features
- Actor-model concurrency with isolated mailboxes and async loops, ensuring no shared state or race conditions
- Runtime agent spawning via LLM-based intent classification, eliminating predefined types and restarts
- MQTT-based messaging for loose coupling, real-time telemetry, and edge-native operation
- Auto-persistence writing agent state to disk on every tick, enabling crash recovery with zero data loss
- Multi-interface support including REST, WebSocket, Discord, WhatsApp, Telegram, and CLI with streaming responses
- Per-agent and aggregate LLM cost tracking across all providers
- Visual drag-and-drop workflow builder with real-time multi-agent collaboration using the AG2 framework
- Deployment options via Docker, systemd, or native Python runner, with SSH-based remote node installation