Repolode
Submit
heymrun

heymrun/heym

Self-hosted runtime and visual platform for building, running, observing, and controlling agentic systems and deterministic workflows.

Open-source alternative to:Zapiern8n

View on GitHubWebsite
Stars
1,371
Forks
114
Open issues
2
Last commit
3 hours ago

What is it?

Heym lets teams compose AI agents and deterministic workflow steps on a visual canvas or generate them from natural language. Workflows connect to LLMs, HTTP APIs, MCP servers, RAG knowledge stores, and common business tools such as Slack, GitHub, and Google Sheets. The runtime executes the graph with parallel branches, retries, and error paths.

Operation features are central to the project. Every execution is saved with inputs, outputs, and node-level state, and LLM traces record tokens, latency, tool calls, and USD cost. Evals compare prompts across models, alerts fire on errors or spend thresholds, and agent review checkpoints pause a run for human approval before it resumes from a saved snapshot.

Workflows can be invoked through REST, SSE streaming, cron and webhook triggers, chat portals, or published as tools for MCP clients. Heym is deployed on your own infrastructure with PostgreSQL and Docker Compose, and it supports multi-instance workers sharing a PostgreSQL queue.

Key features

  • 01Visual canvas and AI generation for agents and deterministic workflows
  • 02Parallel node execution, retries, and error branching
  • 03RAG with Qdrant or PostgreSQL/pgvector and per-agent persistent memory
  • 04Execution history, LLM traces, token and USD cost tracking
  • 05Human-in-the-loop review checkpoints that resume from saved snapshots
  • 06Workflows exposed via REST, SSE, MCP, and chat portals

Use cases

  • Running internal business automations that combine LLM steps with existing SaaS tools
  • Building a support or knowledge-base chatbot backed by company documents
  • Reviewing high-stakes agent actions with a human approval step before they proceed
  • Monitoring model spend and run health for agent workflows in production

Quick start

  1. 1

    Clone the repository

    git clone https://github.com/heymrun/heym.git
  2. 2

    Enter the project folder and start Heym locally

    cd heym && ./run.sh

Things to know

  • The license is MIT with a Commons Clause condition, which restricts selling the software as a commercial product.
  • Running it on your own infrastructure means setting up and maintaining the runtime yourself.

In collections

Open-source AI agent frameworks & appsMCP serversSelf-hosted apps

Topics

#ai-agents#workflow-automation#self-hosted#mcp#rag#llm-observability#human-in-the-loop#no-code#agentic-ai#ai-agents-framework

FAQ

What is heym?
Self-hosted runtime and visual platform for building, running, observing, and controlling agentic systems and deterministic workflows. Heym lets teams compose AI agents and deterministic workflow steps on a visual canvas or generate them from natural language. Workflows connect to LLMs, HTTP APIs, MCP servers, RAG knowledge stores, and common business tools such as Slack, GitHub, and Google Sheets. The runtime executes the graph with parallel branches, retries, and error paths.
Is heym free to use? What license does it have?
heym is open source under the Other license. Check the license terms before using it in commercial products.
Is heym still maintained?
Actively maintained. The last commit to heym was on Nov 7, 2026.
What can I use heym for?
Running internal business automations that combine LLM steps with existing SaaS tools Building a support or knowledge-base chatbot backed by company documents Reviewing high-stakes agent actions with a human approval step before they proceed Monitoring model spend and run health for agent workflows in production
How do I install heym?
Clone the repository: git clone https://github.com/heymrun/heym.git Enter the project folder and start Heym locally: cd heym && ./run.sh
Where can I get heym?
The source code and installation instructions are on GitHub: https://github.com/heymrun/heym

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects