company · about

Built for engineers who ship AI.

tokentrace gives you full visibility into every LLM call your app makes — spend, latency, prompts, and traces — without changing how you write code.

drop-in SDKopen sourceself-hostable

The problem we solve

Most teams flying blind with LLMs. They don't know which prompts cost the most, which model versions introduced a latency regression, or what the exact input/output was when a user reported a bad response.

tokentrace captures every LLM call — input, output, model, tokens, latency, cost — and makes it searchable in a dashboard your team can share.

Who uses tokentrace

solo devsInstrument a side project in under 5 minutes.

startupsCatch cost spikes before they hit your invoice.

ai teamsCompare prompt versions with real production data.

platform engCentralize observability across multiple LLM services.

Supported providers

OpenAI, Anthropic, and Google Gemini — with more providers on the roadmap. One SDK wraps all of them. Switch models; your traces follow automatically.

Open source core

The tokentrace SDK is MIT-licensed and lives on GitHub. You can read exactly what data is collected, fork it, extend it, or self-host the entire platform. We believe observability tooling should be auditable.

github.com/brentibanez331/tokentrace-sdk →