Why We Chose OpenTelemetry for AI Cost Tracking

When we started building botanu, we had a choice: build proprietary instrumentation agents or build on an open standard. We chose OpenTelemetry, and here's why.

The Case Against Proprietary Agents

Proprietary agents create lock-in. They require you to trust a vendor's code running inside your infrastructure. They add another dependency to maintain, update, and debug. And when you want to switch providers, you're looking at a re-instrumentation project.

Why OpenTelemetry

OpenTelemetry is the CNCF's open-source observability framework. It's already adopted by most cloud-native teams for tracing and metrics. By building on OTel, botanu gets several advantages:

Zero new concepts. If your team already uses OpenTelemetry (and most platform teams do), adding botanu is a configuration change, not a new framework to learn.

Vendor neutrality. Your instrumentation isn't locked to botanu. The traces you generate are standard OTel spans that work with any compatible backend.

Minimal footprint. We don't run agents in your infrastructure. Our lightweight Docker collector processes telemetry locally and sends only aggregated metrics to the botanu platform.

Security by design. Because we use standard OTel instrumentation, your prompts, responses, and customer data never leave your environment. Only cost and performance metrics are transmitted.

How It Works in Practice

Adding botanu to a workflow takes three steps:

• Add our OTel-based SDK to instrument your AI workflow spans.
• Deploy our Docker collector container alongside your existing infrastructure.
• Connect your billing APIs (AWS, GCP, Azure, OpenAI, Anthropic).

The collector correlates trace data with billing data locally, then sends aggregated cost-per-outcome metrics to your botanu dashboard. Most teams complete setup in under a day.

The Bottom Line

Open standards mean less risk, less lock-in, and faster adoption. That's why we built botanu on OpenTelemetry.