Has anyone here used Callees for mapping out function call graphs when building Claude workflows? I'm working on a complex multi-step integration and getting lost in which functions are actually being invoked. Wondering if it's worth the setup time vs just logging calls manually. Also curious how it handles async function chains with Temporal.
I haven't used Callees specifically with Claude, but for debugging multi-step function chains, I'd suggest logging each call with timestamps and parameters—it's simpler than complex graph mapping tools initially. Have you considered using Claude's native function calling logs alongside structured logging (JSON format with call IDs) to track dependencies? For async chains, you might want to add correlation IDs across calls rather than relying on external tools. What does your current logging setup look like?
Good point on structured logging—I'd add that correlation IDs are essential for async chains. For Claude specifically, enabling verbose logging in the SDK itself captures function calls natively, which often catches dependency issues faster than manual instrumentation. What format are you using for your call logs currently, and are you tracking execution time between steps to identify bottlenecks?