Agents
See every agent's spend, speed, and health.
Foglamp groups traffic by agent name automatically — so you can see which agents are slow, expensive, or failing, and watch the full call flow for each one.
support-triage · typical run
classify
fetch-order
search-kb
draft-reply
Per-agent rollups
One row per agent. All the numbers.
Requests, error rate, p95 latency, spend, and eval pass rate — for every agentName seen in the last 24 hours. Spot the outlier in a glance.
- Automatic grouping by agent name
- Spend, latency, errors, and eval pass rate per agent
- Which models each agent actually calls
support-triage
6.2k
requests
2.81s
p95
96%
eval pass
Call flow
Watch the agent think, step by step.
Every agent run becomes a flow of LLM calls, tool calls, and sub-agents. See where time and tokens go, and exactly which step failed when something breaks.
- LLM, tool, and sub-agent steps as one flow
- Per-step duration, tokens, and status
- Jump from a flow node into its full trace
classify
fetch-order
search-kb
draft-reply
Your agents are running in the fog.
Cost, latency, errors, eval scores — all there, all invisible. Wrap your model and turn the light on.