Guides, model notes and product updates

TokenLab usage exports now let finance and engineering teams pull matching usage records directly from the dashboard, so spend reviews start with real records instead of guesswork.

TokenLab determines payload schemas by native endpoint selection rather than dynamic format hint headers. Learn how to configure documented headers for billing reconciliation, rate limiting, and task tracking.

TokenLab auto recharge is opt-in, requires a saved Stripe payment method, and fails closed on payment problems. Here is how the trigger logic, defaults, and limits work.

The TokenLab MCP server now covers catalog browsing, paid model and media calls, files, and async tasks, so this guide rewrites our earlier read-only description and walks one discover-check-call flow.

Responses API vs Chat Completions comes down to who owns conversation state, and this guide pairs a decision table with matching requests, tool and streaming differences, per-model format checks, and a migration checklist.

Replicate's per-second billing can turn a quiet week into a surprise invoice. We compare it with TokenLab for teams that need a Replicate alternative AI API.