When you change an agent's model, prompt, tools, or environment, how do you know whether the new version performs better? Production traces show what happens…

When you change an agent's model, prompt, tools, or environment, how do you know whether the new version performs better? Production traces show what happens…

When you change an agent's model, prompt, tools, or environment, how do you know whether the new version performs better? Production traces show what happens…

TypeSafe AI is a model provider that answers typed questions about structured state, returning schema-validated results instead of free-form text.

Ready-to-use evaluation metrics for measuring LLM application quality

Instrument Google ADK for Java with OpenInference and export traces to Phoenix.

Trained for agentic workflows and optimized for competitive coding performance.

The latest from the Phoenix team.

Phoenix's built-in MCP server adds read-only SQL tools and code mode so coding agents can query traces ~17x cheaper than retrieval-only MCP workflows.

PXI completes multi-step UI workflows with approval-gated changes, a new evaluator scores retrievals from any source, and new trace and REST tools speed up analysis and administration.

How Claude Fable 5 and Claude Opus 5 return classifier refusals and how to retry refused requests on a fallback model.