Migration Guides
Moving from another LLM observability tool to TraceCraft is usually straightforward. TraceCraft is vendor-neutral and supports all major export formats (OTLP, JSONL, HTML), so you can keep your existing backend while switching the instrumentation layer.
Each guide covers the key differences, a side-by-side code comparison, a feature mapping table, and a step-by-step migration checklist.
Supported Migrations
Replace LangSmith tracing with TraceCraft. Covers @traceable to @trace_agent
Replace Langfuse SDK calls with TraceCraft decorators and context managers. Covers
Replace OpenLLMetry instrumentation with TraceCraft. Covers span mapping, workflow
Why Migrate to TraceCraft?
| Feature | LangSmith | Langfuse | OpenLLMetry | TraceCraft |
|---|---|---|---|---|
| Vendor-neutral export | No | Partial | Yes | Yes |
| Works fully offline | No | Self-host | Yes | Yes |
| Per-trace pricing | Yes | Yes | No | No |
| Multi-framework support | LangChain | Any | Any | Any |
| Built-in PII redaction | No | No | No | Yes |
| Terminal UI | No | No | No | Yes |
Migration Strategy
Regardless of which tool you are migrating from, the process follows the same pattern:
- Install TraceCraft alongside your existing tool (no need to remove it first).
- Initialize the runtime at application startup with your chosen exporters.
- Replace instrumentation one module at a time - decorators, callback handlers, or context managers depending on your integration point.
- Verify traces appear correctly in the console or TUI before switching exporters.
- Remove the old SDK once all modules have been migrated and verified.
Zero-Downtime Migration
Because TraceCraft can export to the same OTLP backend as your current tool, you can run both side by side during the transition. Simply point both SDKs at the same collector and compare the output before fully cutting over.
Next Steps
Choose the guide that matches your current tool: