telemetryZero
observability + replay for multi-agent systems

The flight recorder &
time machine for AI agent swarms

Multi-agent systems fail silently. telemetryZero traces every tool call, replays any run, diffs the prompt, and lets you fix a failing agent and watch it recover — live.

powered by Tavily web search · OpenAI · Vercel AI SDK

3
agents traced
12+
tool calls / run
0ms
replay latency
0
silent failures

Agent trace viewer

See the full swarm as a tree — every agent, LLM call, and tool hop.

Tool-call timeline

A waterfall of latency, tokens, and cost for each step of the run.

Replay mode

Scrub a playhead through the run and watch decisions unfold live.

Prompt diffing

Red/green diff any two prompt versions to see what actually changed.

Eval scores

LLM-as-judge grades every run for correctness and quality.

⦿

Failure clustering

Group failures — timeouts, bad tool args, hallucinations — into buckets.