agentrace

If you're building AI agents and need visibility into what they're actually doing, this gives you structured trace logging for agent steps, decision points, and errors. It's an MCP server paired with a CLI viewer, so you can instrument your agent workflows and then inspect the execution traces without building your own observability stack. Reach for this when your agent is misbehaving and you need to see the sequence of actions it took, or when you want to understand why it made specific decisions. The combo of MCP integration for data collection and a dedicated CLI for analysis means you can debug agent behavior the same way you'd use logging and tracing for traditional applications.