Migrate to Curate-Me
Every migration guide follows the same pattern:
- Before / after code diff — exactly what changes, nothing else
- Environment variable mapping — old var → new var with explanations
- Auth header changes — how keys and headers shift
- Provider key storage — how to store your upstream secrets
- Cost controls — set a budget from day one
- Rollback steps — revert in under 60 seconds if you need to
Pick the guide that matches your current setup:
What stays the same
- Your LLM SDK (OpenAI, Anthropic, LangChain, etc.) — no code changes
- Your model names — we pass them through unchanged
- Your provider API keys — stored encrypted in Curate-Me; you never lose access
What you gain immediately
- Per-request cost tracking and daily budget caps
- PII scanning on every prompt (33 patterns + optional Presidio NER)
- Rate limiting with IETF-standard
RateLimit-*response headers - HITL approval gates for high-cost or sensitive requests
- A full audit trail in the dashboard
Estimated migration time
| Origin | Typical time |
|---|---|
| LiteLLM | 15–30 minutes |
| Portkey | 10–20 minutes |
| Helicone | 10–20 minutes |
| Langfuse | 20–40 minutes (SDK instrumentation stays in parallel) |
| Direct SDK | 5–10 minutes |
| Self-hosted OpenClaw | 5 minutes |