Sandbox

FOXHOUNDAgent Ops Console

Overview

Fleet Overview Executive Summary

Agents

Investigate

Traces Run Diff Prompts Session Replay

Improve

Datasets Evaluators Experiments

Govern

Budgets SLAs Regressions Notifications

Agent · Sandbox

Refund Policy Chain

refund-policy-chain

Multi-step refund workflow: intent classify → policy lookup → eligibility → response.

View risk posture →

Risk posture

Top driver: SLA

82/ 100 · incident

82 = 45 SLA + 12 Budget + 12 Regressions + 12 Alerts

Version timeline

3 versions

v19canary
model: claude-sonnet-4-6 · deployed 5/23/2026, 8:00:00 AM · created 5/22/2026
Canary in progress · v19
production
10%trafficwebhook 200
Started 5/23/2026, 8:00:00 AM · Initial canary rollout
v18production
model: claude-sonnet-4-6 · deployed 5/12/2026, 10:00:00 AM · created 5/11/2026
v17deprecated
model: claude-sonnet-4-5 · deployed 4/28/2026, 9:00:00 AM · created 4/27/2026

Coming in next sessions

Agent Flow (funnel + path viz) — Session 6
Scoped traces, replay, run diff — Session 7