Phoenix, from Arize AI, gives developers visibility into what their AI agents are doing. It records each step an agent takes, so when something goes wrong you can see exactly where, and it offers tools to measure quality through evaluations and experiments.
Being open source, it can be self-hosted or started locally with a single command. There is also an assistant that lets you ask questions about your traces. It suits AI engineers at startups and larger companies who want to improve agent reliability.
