I am building Phantasm, an open-source tool that allows folks who build AI agents to have human approval layer on top of the agent. With tool-calling becoming more mainstream in agentic AI, I build Phantasm to be a layer of safety allowing humans to guide the AI agent workflow in real-time production.

It is still in early stage of development but I make sure to include components and features that makes it not too far from being production-ready. Phantasm comes with a custom server, dashboard and a Python SDK with multi-approver load balancing for growing teams. I'd love to hear more feedback from the community to make this tool something that brings value to the AI community.

Thank you in advance!

I wrote a guide here to get started with LangChain and Phantasm quickly:

https://dev.to/edwinkys/how-to-build-tools-for-ai-agents-wit...