PawFlow is a self-hosted runtime for AI agents and executable flows. It lets you wire agents, tools, files, relays, services, memory, conversations, and workflows together while keeping tool calls, provider calls, token usage, and failures visible.
No reviews yetBe the first to leave a review for PawFlow
Maker
📌
Hi Product Hunt,
I’m Quentin, solo builder of PawFlow.
I built PawFlow because most agent tools feel like black boxes. They call tools, spend tokens, touch files, fail silently, or get stuck, and you only see the result at the end.
PawFlow is built around the opposite idea: agents, tools, files, relays, services, memory, conversations, and workflows should be visible, traceable, and composable.
It is inspired by Apache NiFi, but built for AI agents.
I’d love feedback from builders working with agents, automation, self-hosting, and developer tools.
Report
How does the self-hosted setup handle scaling when agent workloads spike, and is there any built-in observability for token usage across providers or do I need to wire that up myself?
Report
The self-hosted setup was smoother than I expected, and being able to watch token usage and tool calls live made debugging way less painful.
How does the self-hosted setup handle scaling when agent workloads spike, and is there any built-in observability for token usage across providers or do I need to wire that up myself?
The self-hosted setup was smoother than I expected, and being able to watch token usage and tool calls live made debugging way less painful.