I built Xpectrum to make AI agent workflows easier to build and run.

Hey PH community! 👋

I'm building Xpectrum AI and just launched it on Product Hunt.

I've been experimenting with AI agents and kept running into the same problem: once an agent needs to use tools, retrieve information, follow multiple steps, or wait for human approval, the workflow quickly becomes a combination of different frameworks, APIs and custom code.

So I started building Xpectrum around that problem.

What's inside:

• AI agents — build agents that can reason over a task and take actions
• Knowledge base — connect documents and use them as context with retrieval
• Tool integrations — let agents work with APIs and external tools
• Workflow builder — create multi-step agent workflows
• Human approval — pause workflows when a person needs to review an action
• Observability — see what the agent did and where a workflow stopped

The idea is pretty simple:

Agent + Knowledge + Tools + Workflow + Human approval

Instead of building every part separately.

I'm still working on it and there are definitely things that need improvement.

Would love to hear from other builders:

1. What is the biggest pain point you've faced when building AI agent workflows?

2. If you're using n8n, Zapier, CrewAI or custom code, what do you wish was easier?

Would really appreciate any feedback from the PH community.