Helicone doesn't give me enough detail on tool calls, what should I switch to?

If you need much richer visibility into tool calls than Helicone is giving you, the main “switch to” options are:

1) Langfuse

Best all-around choice for LLM tracing + tool/function call observability.

Why it’s a good fit:

• Shows full traces across agent steps
• Captures tool calls, inputs, outputs, latency, errors
• Good support for nested spans and multi-step workflows
• Open source + hosted options
• Easier to inspect agent behavior than simple request logging

2) OpenTelemetry + your own backend

Best if you want maximum control and already have an observability stack.

Why:

• You can instrument tool calls as spans/events exactly how you want
• Works well if you already use Grafana/Tempo/Jaeger/Datadog
• More setup, but far more flexible

3) Datadog / Honeycomb

Best for teams that want production-grade observability and already use those tools.

Why:

• Excellent distributed tracing
• Strong filtering/search over metadata
• Good for debugging tool-heavy agents in production
• More expensive, and less “LLM-native” out of the box than Langfuse

4) LangSmith

Best if you’re in the LangChain ecosystem.

Why:

• Very good trace inspection
• Strong agent/tool visibility
• Great for LangChain-based apps
• Less ideal if you’re not using LangChain

---

My short recommendation

• If you want the best replacement for Helicone for tool-call detail: Langfuse
• If you want enterprise observability and already use a tracing platform: Datadog or Honeycomb
• If you want full control: OpenTelemetry

If you want, I can also give you a direct feature comparison of Helicone vs Langfuse vs LangSmith specifically for tool call tracing.