Kevin D.

This looks great.... recognize the ClaudeSDKClient that drives both Claude Code and the Agent SDK 🙂 Only issue in my case is the LiteLLM requirement. I'm taking advantage of the GLM-4.6 Plan ($) to drive CC and also my agent system so would be an issue to drive with LiteLLM. Will check it out.. Thanks for the quick response. This new CC approach of the coding agent "taking more control" of other Agent Systems and processes, I'd imagine that observability tracing is going to be really in demand... the callback system they have for getting ANY internal view logs is a nightmare.

Hi Aman. I want to use it to interpret the trace results post run.... such as staying in the IDE with AI to help with agent development. You know most of the important stuff for agent debugging dev is in the agent black boxes, not the python. I already use the new Arize Tracing Assistant mcp to guide Windsurf etc to teach my IDE to better instrument tracing.. that was very helpful. Now im selfishly looking for another tool. I've used a similar mcp (from that unnamed... sounds like "bogfryer") to do this for pydantic AI agents. In that case, i was originally using their API to get custom analysis through python runs... but their mcp was a game changer as it lets me do ad hoc questioning without having to leave IDE, generate reports, etc. Since i saw you guys starting to integrate with gemini and ADK, i wanted to try that agent dev through Arize.. thanks!