When AI agents start drifting, how do you debug them?

Hey Product Hunt! 👋

We’ve been running agent workflows for years, and one problem keeps coming back.

It’s rarely that an agent can’t do the task. It’s that, over time, things start to drift.

One agent makes a wrong assumption, another takes it as fact, and the error compounds and soon you’re debugging a chain of decisions you can barely see.

The usual response is to add more instructions and crossing fingers. But without seeing what actually happened, it's a stab in the dark.

That’s why we built ⁠guilds.run: open-source AI agent coordination framework, where you can read and edit prompts and memory, inspect every tool call and result, follow conversations between agents, control permissions, and choose the model for each agent.

We’re launching today and If this problem sounds familiar, we’d love your support 🫶