The new telemetry.dev version

telemetry.dev is a platform that enables observability for AI applications. Once integrated, all model calls, tool usage, and retrieval steps within an agent’s execution process are linked together into a complete chain. It is possible to clearly see how much time each step takes, how many tokens are consumed, what costs are incurred, and where any errors occur.

It is built on OpenTelemetry and provides SDKs for TypeScript and Python. Projects that are already using OTLP can integrate it directly. In addition to link tracing, it also offers MCP traffic analysis, cost and error exception alerts, prompt version management, as well as evaluation records.

The free package offers 30,000 data units per month and does not require the use of a credit card. It is ideal for development teams that have already deployed their intelligent agents and are now concerned with costs and stability.