Introduction
DesignOperational telemetry for AI systems: connect the run, expose the expensive span, carry a concrete action back.
#What Telemeter is
Telemeter collects telemetry from production LLM and agent applications, connects it under one stable run identity, and turns the result into a concrete next action — delivered where your team already works. Logs tell you what happened; Telemeter is built to tell you what to do about it.
It is OpenTelemetry-native, EU-hosted by design, and Apache-2.0 open core with a documented self-hosting path.
#The shape of the product
- Ingest — bounded OpenTelemetry payloads, unsafe shapes rejected at the edge.
- Connect — model calls, tools, retries, cost, latency, and outcomes under one run.
- Analyze — read-only agents reason over the spans that actually ran.
- Act — recommendations arrive with the evidence that produced them, in Console, email, or MCP.
The public site and free tools are live today. The ingestion API, agents, and MCP runtime are designed and documented here, but not yet built.