The Heym Platform

Compare Heym to traditional automation tools, explore production AI use cases, review the tech stack, and learn how to self-host with Docker Compose or Kubernetes.

Why Heym?

See how Heym compares to other automation platforms. Built AI-first, not AI-added.

FeatureHeymn8nZapierMake
Built-in LLM NodeSend prompts to language models for text generation, vision, image creation, and structured JSON output.
LLM Batch API + Status BranchesSend an array of prompts through the OpenAI Batch API, with a dedicated status branch for live progress from pending through completed—alongside your main result path on the canvas.
Built-in Agent Node (Tool Calling)LLM-powered agents that can call Python tools, execute code, and interact with external services autonomously.
Multi-Agent OrchestrationOne orchestrator agent delegates tasks to named sub-agents and sub-workflows with up to 5 levels of nesting.
Built-in RAG / Vector StoreUpload documents to managed Qdrant vector stores and perform semantic search with metadata filters and reranking.
WebSocket Read / WriteNative inbound and outbound WebSocket workflow steps for listening to external streams and pushing realtime payloads.
Natural Language Workflow BuilderDescribe what you want in plain text or voice and the AI assistant generates the entire workflow on the canvas.
MCP (Model Context Protocol)Connect agents to MCP tool servers or expose your workflows as an MCP server for Claude Desktop and Cursor.
Skills System for AgentsPortable capability bundles with SKILL.md instructions and optional Python tools that extend agent behavior via drag-and-drop.
LLM Trace InspectionView full request and response payloads, timing breakdowns, tool calls, and skills included for every LLM invocation.
Built-in Evals for AI WorkflowsCreate evaluation suites with test cases, run them against multiple models, and compare pass/fail rates with LLM-as-Judge scoring.
Human-in-the-Loop (HITL)Agents pause at approval checkpoints, generate a public review link, and wait for a reviewer to accept, edit, or refuse.
LLM GuardrailsBlock unsafe content categories like violence, hate speech, or harassment with configurable sensitivity levels per node.
Automatic Context CompressionAgent conversations automatically compress when reaching context limits to prevent timeouts and enable long-running tasks.
Parallel DAG ExecutionThe engine builds a directed acyclic graph and runs independent nodes concurrently with a thread pool for maximum throughput.
Self-Hostable, Source AvailableDeploy on your own infrastructure with Docker Compose or Kubernetes. Your data never leaves your servers.
Expression DSL for Dynamic DataReference upstream node outputs with expressions like $nodeLabel.field, with support for arithmetic, string helpers, and array operations.
Native, first-party support
Partial or plan-dependent
No documented native support

Comparison reflects publicly documented, native product capabilities reviewed on April 21, 2026. Availability may vary by plan, deployment model, or third-party integrations. “Partial” indicates limited, plan-restricted, or indirect implementation — hover for details.

Heym was designed AI-first from day one — multi-agent orchestration, built-in RAG, a portable skills system, and parallel DAG execution are core primitives, not add-ons. Combined with full self-hosting, LLM trace inspection, and a natural-language workflow builder, Heym gives teams complete control over their AI automation stack without vendor lock-in.

Built for Every Team

From customer support to DevOps, Heym adapts to your workflow needs with AI-powered automation.

Each use case below shows a real workflow pattern you can build on the Heym canvas. Combine trigger nodes for scheduling or webhooks, AI nodes for language understanding and generation, logic nodes for branching and looping, and integration nodes for connecting to external services. Every workflow runs as a parallel DAG with automatic dependency resolution, so steps that can execute concurrently do so without extra configuration.

Automate Customer Support with AI Agents

Build intelligent support workflows that understand context, search knowledge bases, and resolve issues autonomously. An Agent node retrieves answers from your RAG-powered documentation, applies guardrails for safe responses, and escalates to a human reviewer through the built-in approval checkpoint when confidence is low.

  • RAG-powered knowledge base with Qdrant vector stores for instant semantic answers
  • Human-in-the-loop escalation generates a public review link for complex cases
  • Multi-language support with configurable guardrails to block unsafe content
  • IMAP inbox polling plus Slack and email delivery for omnichannel support triage
  • Portal mode turns the workflow into a public chat UI for end-user self-service
Try This Workflow

Customer Support Workflow

7 nodes · Powered by Heym

IMAP Trigger
Qdrant RAG
AI Agent
Guardrails
HITL Review
Slack Notify
Send Email

Built with Modern Tech

Every component is chosen for performance, developer experience, and reliability.

Python
TypeScript
Vue.js
FastAPI

Frontend

The application frontend is built with Vue.js 3 and TypeScript for type safety, with Vue Flow powering the visual canvas editor. Vite and Bun provide fast builds, while Tailwind CSS and Shadcn Vue handle styling.

Vue.js 3
Framework
TypeScript
Language
Vite + Bun
Build
Vue Flow
Canvas
Pinia
State
Tailwind CSS
Styling
Shadcn Vue
UI

Backend

A Python backend powered by FastAPI delivers async performance for concurrent workflow executions. UV manages packages and Alembic handles database schema migrations.

Python 3.11+
Language
FastAPI
Framework
UV
Package Mgr
Alembic
Migrations

Database & Infra

PostgreSQL stores workflows and execution history, Redis handles caching and rate limiting, RabbitMQ provides message-driven triggers, and Qdrant powers vector search for RAG pipelines.

PostgreSQL 16
Database
SQLAlchemy 2.0
ORM
Redis
Cache
RabbitMQ
Queue
Docker
Container
QDrant
Vector DB

Security & Auth

JWT tokens in HttpOnly cookies with refresh token rotation secure user sessions. Passwords use bcrypt hashing, credentials are encrypted at rest with AES-256 Fernet, and Pydantic v2 validates all inputs.

JWT
Auth
bcrypt
Passwords
Pydantic v2
Validation
Fernet
Encryption

AI & LLM

Connect to OpenAI, Ollama for local models, vLLM for high-throughput inference, or Cohere for embeddings and reranking. MCP support lets agents connect to external tool servers or expose workflows as tools.

OpenAI
LLM
Ollama
Local
vLLM
Inference
MCP
Protocol
Cohere
LLM

Developer Experience

First-party documentation lives inside Heym—structured guides for every node, tab, and reference topic. Developers and AI agents get a native chat-with-docs experience to explore and apply the docs without leaving the product. The codebase is kept healthy with ESLint and Ruff lint checks and pytest-backed unit tests.

In-app documentation
Native
Chat with docs
Built-in
Agent-friendly docs
Agents