The team behind Heym
Two co-founders and engineers building an AI workflow platform for teams that need control of their data, agents, and production runs.
Heym (also known as Heym AI) is a source-available, self-hosted AI workflow automation platform at heym.run. Build multi-agent AI pipelines, retrieval-augmented generation workflows, and Model Context Protocol connected automations on a visual canvas.
Why we built Heym
We built Heym because teams should not have to choose between rapid AI experimentation and operational control. Most workflow tools excel at fixed automation, but production AI systems need agents that can use tools, retrieve knowledge, hand work to other agents, and pause for approval.
Heym makes that work visible. The visual canvas keeps deterministic steps explicit while giving agents the tools they need to handle judgment and context. Built-in MCP, retrieval, human review, and execution history keep the workflow around the model as deliberate as the model itself.
We built Heym as a source-available, self-hosted platform because control matters when workflows touch real data and real decisions. We ship it in public for builders who want to inspect, adapt, and improve their AI systems over time.
What we've built
Heym brings the building blocks for AI workflow automation into one platform. These are the capabilities we use ourselves and continue to develop.
Visual workflow canvas
Design AI workflows from a prompt or assemble them on a visual canvas. Combine triggers, logic, code, integrations, browser automation, and data steps, then test with pinned data and live node output.
Agents and tool calling
Give agents controlled access to canvas nodes, Python tools, skills, external MCP servers, and other workflows. Use structured output, context management, and delegation to build reliable single-agent or multi-agent systems.
Human review and control
Pause a run for review before an important action. Reviewers can approve, edit, or decline an output, and the workflow resumes from the saved state. Send review links through Slack, email, or any workflow.
RAG and vector search
Build retrieval-augmented generation workflows with Qdrant or Postgres and pgvector. Upload documents, filter by metadata, use reranking when needed, and pass grounded context to your agents.
Evals, LLM traces, and alerts
Run test suites across models, score outputs, and compare changes before release. Inspect LLM traces, node timing, and tool calls, then create alerts for failures, slow runs, or spend.
Boards, dashboards, portals, and MCP
Turn workflow results into dashboards, run agentic work through boards, publish portal experiences, or expose workflows as MCP tools for clients such as ChatGPT, Claude, and Cursor.
Built for production
Deploy Heym on infrastructure you control. The stack includes PostgreSQL, a FastAPI backend, and a Vite frontend. Credentials are encrypted at rest, and workflow webhooks support JWT or custom header authentication.
Teams can share workflows, credential access, vector stores, and global variables without distributing raw API keys. The Global Variable Store persists useful state across runs, from counters and accumulated outputs to agent context.
The execution engine runs independent nodes concurrently and records every run with its inputs, outputs, node results, and timing. Execution history makes it easier to debug a workflow, understand what happened, and run it again with the same inputs.
How we write
We publish a lot about the software we build, so it matters how those articles are put together. Every post carries a named author, is reviewed by the other founder before it goes live, and states its sources. Claims we make about Heym are checked against the code rather than written from memory, and when we get something wrong we fix the article rather than quietly leaving it. The full rules, including how we use AI assistance and how we handle corrections, are in our editorial policy.
Team
Ceren Kaya Akgün
Co-founder & Engineer
Ceren Kaya Akgün is a co-founder and engineer at Heym with a background in distributed systems, artificial intelligence, and machine learning. She led the design of Heym's visual workflow canvas and agent orchestration engine. Over the past several years, she has built AI systems that operate reliably at scale. She writes about turning AI research into dependable products, with a focus on the tools, tradeoffs, and architecture decisions that separate a demo from a production system.
Mehmet Burak Akgün
Co-founder & Engineer
Mehmet Burak Akgün is a co-founder and engineer at Heym with a background in backend engineering and platform infrastructure. He designed and built Heym's workflow execution engine, live streaming architecture, and deployment system based on Docker. His work focuses on making powerful AI automation accessible to teams that want full control of their data and infrastructure without relying on a single vendor or sending data outside their own stack.