ByteChef
Build AI Agents & Automate Workflows, Open Source
ByteChef is an open-source automation platform for building workflows, multi-step AI agents, and integrations between the apps a business already runs on.
We started it because automation today forces an uncomfortable trade-off. Hosted platforms are easy to start with, but you end up locked into someone else's runtime, paying per task as you scale, and handing your operational data to a third party. The self-hosted alternatives give you control but usually stop at moving data from A to B.
The part we care about most is AI agents. Chaining a single LLM call into a workflow is easy. Building an agent that reasons over your own systems, calls your own tools, and runs reliably in production is not. We want that to be within reach of a small team that doesn't have a platform engineering group behind it.
Technical teams come for the flexibility. Because it is open source and self-hostable, developers can run it on their own infrastructure, extend it with custom components and code, wire agents into internal APIs, and treat workflows like any other versioned part of the stack. Business users get a visual builder, engineers get something they can open up and change, and both work on the same platform. Most tools make you pick one.
Where things stand: the platform is in beta and open source on GitHub, with a growing integration library covering the usual suspects such as Gmail, Slack, Salesforce, HubSpot, Jira, Postgres and OpenAI. It's aimed at small and medium businesses priced out of enterprise automation tooling, and at the developers building for them.
Next up: deeper agent tooling, more integrations, and self-hosting that is genuinely easy to run. Happy to answer questions about the open-source model, the agent side of things, or anything else.
5 Comments
The self-hosted angle makes this more interesting than another visual automation builder. The real value seems to be giving teams control over the runtime and data while still making agent workflows accessible without a platform team.
Curious how you’re thinking about the line between “easy enough for business users” and the flexibility engineers expect from a self-hosted system.
We see both of them as a collaboration. We are exposing, let's say, lower layers of the platform (the ability to write code) to give them full flexibility, and for business users we have a no-code/low-code approach through the visual builder. But everything is sitting in the same infrastructure.
That makes sense. I’d be interested to hear what the first users reveal about that balance. If you’re open to it, what’s the best email to reach you on?
This is an interesting direction, especially the distinction between simply connecting apps and building agents that can reliably reason over real business systems.
The lock-in problem becomes even more important with AI agents because the runtime isn't just executing workflows—it increasingly becomes part of the engineering architecture.
I'm exploring a similar problem from another layer with NAEOS: how to give AI agents a vendor-neutral engineering environment with explicit standards, workflows, governance, and quality gates rather than letting each agent improvise its own way of working.
I think open-source agent infrastructure and an open engineering framework could complement each other really well.
Curious to see how ByteChef approaches agent reliability and observability as the workflows become more autonomous. That's where I think the real challenge begins.
We will be adding auto memory and more advanced session memory handlers to allow agents to be more autonomous.