ByteChef — AI agents and workflow automation

★ 1K

ByteChef is an open-source platform for AI agents and workflow automation that you can self-host or embed in your own SaaS; it positions itself as an alternative to n8n and Zapier. Reach for it when you need to connect services into one automated process, when an AI agent should be a step in that process and, after a failure or a pause for a human, continue from that step instead of restarting, or when data must not go to third-party clouds. Inside: a drag-and-drop visual editor, an AI Agent component with streaming and structured output, a Copilot that builds workflows from a sentence and explains failed runs, 14 LLM providers (OpenAI, Anthropic, Azure OpenAI, Bedrock, Mistral, Groq, DeepSeek, Ollama and more), 8 memory backends, RAG with 15+ vector stores, 12 guardrail types, every component usable as an agent tool, and MCP in both directions. It is written in Java and runs in Docker (Docker Desktop). The embedded H2 database is for evaluation only; the README recommends Docker Compose with PostgreSQL for real use, and the knowledge base and Copilot need pgvector. The core is Apache 2.0. The direction is agent orchestration inside processes, not plain chat. Not a fit if you only need a simple connection between two services without agent logic.