tok: o200k_base · BPEp50 — · —PSPriya S. · Platform Eng
Every prompt has a cost. Krindel makes it the right one.
The intelligence layer between your people and every AI tool — cache first, cascade to the cheapest capable model, cap the spend.
or try one yourself
ExtensionDLP armed0 tok · direct — → — est
Enter to send · Shift+Enter newline · 1–5 load scenarios · cache-first, sanitised, spend-capped before egress
EGRESS BLOCKED
Krindel governance intercepted this prompt in-browser. Nothing left the device.
SPEND CAP REACHED
Policy KRN-BUD-7 held this request before dispatch — the workspace hard cap would be exceeded.
HOLD-2026-0000ws_meridian_eu1finops notified
Model price book
Every price here is a published list rate; costs are computed live from this table using an embedded o200k_base BPE tokenizer, with output length estimated per request. Nothing is faked.
Rates are $/1M tokens (input / output), editable in the PRICING map in-file. Tier 1 = cascade targets · Tier 2 = frontier/baseline.
Gateway settings
Sandbox synthesises upstream responses locally — full pipeline, zero external calls.
Held in memory only — never persisted. Cache, cascade and DLP still run in-browser first; the routed model receives the optimised payload.
list prices · USD
Model list prices are quoted in USD. Any other currency applies this rate, which is shown alongside every converted figure. Update it before pitching.
Drives the pace line on the savings hero.
New here?
Press ▶ Watch the tour — or pick a scenario below it.
Watch the pipeline run in Cascade routing on the right.
Read the savings at the top. That is the whole pitch.