Build Agents
Agent Observability Platform
OpenTelemetry-compatible tracing, real-time cost tracking, and LLM-powered anomaly detection for multi-agent AI systems.
The problem
Once a workflow spans multiple agents calling tools and each other, teams lose visibility into where latency, cost, and failures actually happen.
- Something's slow, but you can't tell if it's the planner, the tool call, or the model.
- The bill went up and nobody can say which agent is burning the budget.
- Debugging means reading raw logs from four different services side by side.
What NEO built
NEO built an OpenTelemetry-compatible instrumentation SDK that ties every agent step to a shared trace ID and exports spans to any OTLP-compatible backend.
The result
Self-hosted, full session replay
Makes slow paths and failure clusters visible across a swarm, with side-by-side comparison between agent runs and releases.
From the blog · 8 min
Multi-Agent Observability: Tracing, Cost Tracking, and Anomaly Detection for AI Workflows
NEO built a production-ready observability platform for multi-agent AI systems, featuring OpenTelemetry-compatible tracing, real-time cost tracking, LLM-powered anomaly detection, and self-hosted deployment.
Try this in your workspace
Paste this into NEO chat to kick off the same workflow on your own data.
Instrument my multi-agent workflow with OpenTelemetry so every agent step shares a trace ID, and export the spans so I can see where latency and cost actually go.
Paste it in · review the plan · get the diff
Get NEO
