Analytics
A single dashboard for every API call across every agent — usage, latency, cost, and errors aggregated daily and broken down per agent.
What you see
Analytics · app.sentientone.ai/analytics
Analytics — last 7 daysAgent: AllRange: 7dExport CSV
Total tokens
1,243,508
+12% vs prev
Avg latency
1.42 s
−180 ms
Requests
4,210
+8%
Total cost
$18.40
−$2.10
MonTueWedThuFriSatSun
| Agent | Requests | Tokens | Avg latency | Cost |
|---|---|---|---|---|
| Website Chatbot | 1,840 | 612,300 | 1.1 s | $7.20 |
| Order Status Agent | 920 | 210,400 | 1.6 s | $4.10 |
| Customer Support | 780 | 284,100 | 1.4 s | $4.90 |
| HR Assistant | 420 | 98,200 | 1.2 s | $1.30 |
| Product Lookup | 250 | 38,508 | 0.9 s | $0.90 |
Usage 42% of plan23 days remainErrors: 12 (0.3%)
Stat tiles
- Total tokensCombined input + output tokens sent to the LLM. A spike in tokens without a spike in requests usually means longer prompts or larger replies.
- Avg latencyEnd-to-end from request start to final response, including any MCP, knowledge, and web-fetch round trips. Sustained >3s points at a slow MCP server or a mismatched model.
- Current usagePlatform requests against your plan quota. Bar turns amber at 75%, red at 90% — upgrade before agents start failing.
- Total costEstimated LLM spend using each agent's configured Token Cost rates. Keep rates accurate to match your provider bill.
- Total requestsDaily volume of API calls. Sort by day to spot weekend troughs or campaign-driven peaks.
- ErrorsCount of 4xx/5xx responses including upstream LLM rate limits, content-policy blocks, and model errors. A spike usually surfaces in Tracing first.
Charts and tables
- Daily volume chartStacked bars showing requests per agent per day. Hover any bar to see exact counts; click an agent in the legend to isolate it.
- Per-agent breakdownTable of requests, tokens, latency, and cost per agent. Sort any column to find your noisiest agent or slowest responder.
- Date-range pickerBuilt-in shortcuts cover today, 7 days, 30 days, and 90 days. Chart granularity auto-adjusts — daily for short windows, weekly for longer ones.
- Agent filterScopes every tile, chart, and table to a single agent. Useful when allocating cost per team or use case.
Use it well
- 1
Daily health check
Open with the 7-day window and the agent filter set toAll. Confirm tokens, requests, and error count look normal — anything off-trend is your starting point. - 2
Drill into one agent
Pick the agent showing the anomaly. Stat tiles narrow to that agent so you can compare against its normal baseline. - 3
Jump to a single request
From the breakdown table, click an agent's name to open Tracing with the same filter applied. You'll see the specific request that drove the trend. - 4
Read the Recommendations
Below the breakdown, opinionated guides cover cost-cutting, prompt tightening, and securing your agents. Pick the one that matches the anomaly above.
Cost accuracy
Cost is computed from each agent's Token Cost tab. If the dashboard cost looks wildly off from your provider bill, update the per-million-token rates there — the next aggregation run picks them up.