OopsTrace Blog

Insights on
AI observability.

Deep dives, tutorials, and product updates from the OopsTrace team.

Introducing OopsTrace 1.0: live trace streaming for AI agents
We're launching OopsTrace out of beta. Here's everything that's new — real-time span streaming, the AI fix assistant, multi-agent support, and more.
Tracing across agent boundaries: how we built the multi-agent span tree
When agent A spawns agent B which calls a tool, how do you represent that in a single trace? We share the architecture decisions behind our hierarchical span model.
Prompt version tracking: catch regressions before your users do
Changing a system prompt mid-deployment can silently break downstream outputs. Learn how to tag prompt versions in OopsTrace and compare outputs across traces.
The 5 most common AI agent failure modes (and how to spot them in traces)
From infinite tool loops to context window overflows, we've analyzed thousands of agent traces. Here are the failure patterns that show up again and again.
How we stream 50k spans/second with zero dropped events
Building a real-time trace pipeline that's both fast and reliable is harder than it looks. A deep dive into our WebSocket architecture and backpressure strategy.
Instrument your CrewAI agents in under 2 minutes
CrewAI's multi-agent orchestration generates complex execution graphs. Here's how to add OopsTrace in a single decorator and get full visibility instantly.

Stay in the loop.

Get new posts, tutorials, and product updates delivered to your inbox. No spam, unsubscribe any time.