1 · What data do you have?
2 · Endpoint / model
3 · Input data
4 · Send
Results
Input report (how codes mapped to features)
Explanation (top contributing features)
Raw REST response
Raw MCP response
Run-all summary
Response timing per endpoint (this browser, all sessions)
Client RTT = full browser round trip. Server = sum of server-side feature mapping + inference (from the response). Network/overhead = RTT − server.
Latency over time (last 200 calls)
Call history (last 500, stored locally)
Key / account usage (from /v1/me)
Per model (billable hits)
Per endpoint (requests)
Daily
Targets (API base URL / IP)
Point at any running stack: http://127.0.0.1:8080, a LAN/Tailscale IP such as http://100.x.y.z:8080, or a public https://api.example.com. When this console is served by its own container, /api proxies to the bundled API.
API key vault (stored in this browser only)
Keys are sent as X-API-Key on REST and MCP calls. The admin console creates keys and sets limits.