Inside the Agent SDK: How we debugged the Claude Agent SDK

Inside the Agent SDK: How we debugged the Claude Agent SDK

About the session

Ahmed Bashir and Ritesh Goru share how the team built and debugged DevRev's Agent SDK on top of the Claude Agent SDK. When the TypeScript implementation did not expose enough runtime visibility, they used request interception, reverse proxies, and MITM-style tracing to inspect behavior and reduce guesswork.

This developer-focused session explores practical techniques for working with agent frameworks whose documentation and public interfaces do not reveal the whole picture. The lessons extend beyond a single SDK to teams building integrations on top of agentic infrastructure.

Key takeaways

  • Understand how DevRev layers its Agent SDK on top of existing agent frameworks.
  • Recognize how limited runtime visibility complicates debugging.
  • Explore request interception, reverse proxying, and MITM-style inspection.
  • Apply a more systematic approach to diagnosing agent SDK behavior.

Agenda

  1. Agent SDK architecture

    Building on the Claude Agent SDK and other frameworks.

  2. Debugging with limited visibility

    Why the public interface and documentation may not be enough.

  3. Traffic inspection techniques

    Request interception, reverse proxies, and MITM-style tracing.

  4. Lessons for agentic integrations

    Reducing guesswork and building with greater confidence.

Speakers

  • Ahmed Bashir

    Ahmed Bashir

    Chief Technology Officer, DevRev

  • RG

    Ritesh Goru