SpendLens AI docs
Choose your runtime. Find your first AI cost leak.
Use Python or Node.js to connect OpenAI and Anthropic spend to the workloads in your code without proxying model traffic.
Quickstart
Choose Python or Node.js, create a key, and send your first tracked call.
Read guide →Python SDK
Use @observe, tracked clients, endpoint tags, and privacy modes.
Read guide →Node.js SDK
Use observe(), async-safe tags, OpenAI Responses, and Anthropic Messages.
Read guide →Compare Mode
Test three policy-approved models safely in staging before switching.
Read guide →OpenTelemetry
Coming soon: send standard AI telemetry from Java, Go, .NET, and other services.
Read guide →API keys
Understand test and live keys, one-time display, and revocation.
Read guide →Privacy
See exactly which operational metadata SpendLens stores.
Read guide →Privacy by default
SpendLens never stores full user or customer prompt content. Template-only mode may capture system instructions or reusable prompt templates; metadata-only mode captures no prompt text.