Heym

Heym is an infrastructure platform designed for building, running, and governing AI agents in production environments. It features a visual editor that allows users to construct agentic systems by connecting a diverse library of node types, including triggers, AI, logic, data, integrations, automation, and utilities. Users can describe a workflow in natural language, and an AI assistant will generate it, or they can manually wire it from a deep library of built-in nodes and integrations.

Core Functionality

The platform supports various node types, such as LLM nodes for text generation and vision, Agent nodes with tool calling and Python tools, and RAG for semantic search over Qdrant or built-in Postgres (pgvector). It also integrates with the MCP client and server, enabling agents to connect to external MCP servers and expose published workflows as callable tools.

Key Features for Production

• **Human-in-the-Loop (HITL):** Includes approval checkpoints and content guardrails to keep humans in control of critical actions.
• **Parallel Execution:** Workflows are compiled and executed as directed acyclic graphs (DAGs), allowing for parallel processing where dependencies permit.
• **Skills System:** Offers portable, reusable capability bundles for Agent nodes, allowing agents to extend their system context and toolbox without modifying node configurations.
• **Browser Automation:** Features Playwright browser automation with auto-heal capabilities.
• **Observability:** Provides live operational dashboards, AI agent observability, cost evaluation, release gating, monitoring, and alerting.

Heym is source-available and fully self-hosted, ensuring that data, agents, and runs remain on the user's controlled infrastructure. It can be deployed with Docker Compose or Kubernetes.